CA2193381A1 - A smooth rain-responsive wiper control - Google Patents

Abstract

An automatic windshield wiper control system responds to sensed moisture in a manner which is responsive to changes in conditions. Moisture (40) and noise (43) sensitive signals are digitized (11), and a functional block (42) detects the reversals of the moisture sensing signal towards a quiescent level. These reversals are considered rain events and on ongoing measurement of the frequencies of these events are produced. The resulting signal is representative of the flow rate of raindrops impinging upon the moisture sensor. A resulting rain intensity signal (49) permits rapid but bounded response to sudden changes in conditions and an appropriate wiper actuation rate (73) is determined. The control strategy features hysteresis, giving the system the propensity to continue operating in a given mode.

Description

096~7026 j ~ 2 1 9 3 3 8 1 PCT~S96/06102 A S~OOTH RAIN-K~o..~lY~ WIPER CONTROL

~riK(J~I) OF ~ lNv~ l.,N
I. Field of the Invention The present invention describes a rain-responsive wiper control system which ~e~ol.ds rapidly to changes in moisture conditions, yet does not change modes of operation so frequently as to be subjectively erratic. As a result, the resulting system is simul~nPollcly smooth and responsive--two characteristics not available in prior art wiper control systems.

II. Dic~llccion of th~ Prior ~rt A moisture sensing win~chipld wiper control system must neco~c~nily employ some moisture sensing means.
Referring to Figure 1 of the drawings, this sensing means may, for purposes of illustration, be an optical sensor 1, such as is disclosed in the ~ r et al. Patent 4,620,141 and the Teder Patent 5,059,877 ~icpoRe~ on the inside surface of a win~chi~l~ 2, within the path 3 swept by wiper blades 4. This moisture sensing means 1 may also be capable of sensing di~l rl~n~$, such as shadows, as described in U.S. Patent Number 5,059,877. For further purposes of illustration, the F of a moisture-sensitive wiper control system may be partitioned into the functional elements illu~LL~ted in Figure 2. In Figure 2, the output lO of a rain sensing means 1 is coupled to the input of an analog-to-digital con~erter ll. A smoothing algorithm 12, implemented either in hardware or software, then actuates the wipers 4 in what it determi n~C to be an optimum manner, by applying d~ Liate signals to a wiper motor 13, by way of a vehicle interface 14. Input from W096l37026 ~1 9338 l Ec./~ 8C~I02 the driver cnnCorn; ng operating mode and desired system sensitivity is imparted to the vehicle interface 14 by means of a driver ~ccoccihle switch 15, convont;on~lly mounted on the steering column. A suitable vehicle interface has been ~;R~OCO~ in U.S. Patent 5,239,244, which is also ~cRignod to applicant~s ~RRignoe.
In its most primitive form, the smoothing algorithm 12 may simply run the wiper motor 13 when the yL~sence of moisture iB detected, and many prior art rain-responsive o wiper control systems posit this as a means of control (e.g. Noack, Us patent 4,355,271). This simple method suffers at least two drawbacks: 1) The area of the W; n~ch; ol d sampled by the sensor is small compared to the windshield as a whole, thus tending toward erratic behavior arising from the random nature of the signal:
and, 2) Even were the sample area of the w;n~ch~ol~ large enough, it i5 subjectively annoying to the driver of t_e vehicle for the wiper to actuate in an erratic manner, even if the rainfall itself is fluctuating in an erratic manner.
Thus some degree of moi~LuL~ sen60L ~ ..se smoothing is desirable. One method, realized with analog electronic -nts, is disclosed in the above-referenced U.S. Patent 4,620,141. In U.S. Patent 5,059,877, the smoothing function is refined and implemented in the software of a mi~Lvc~ L~ller.
Acceptable system behavior is obtained from this method, but the resulting performance constitutes a tradeoff between smoothness and responsiveness.
Figure 3 diagrammatically illustrates the prior smoothing method e '~ in the '877 patent. Sensing means 1 produces a signal on line 10. It is coupled to a ~ ~ 096/37026 ~-~ CP ~~ 2 ~ 9338 ~ F~ll~ 6~6102 block 21 which takes the absolute value of the deviation Or the signal 20 from its ~lioccont- level. The resulting rain deviation signal 22 is coupled to a curve shaping means 23 which produces pulses 24 which are proportional to the degree to which the deviat.ion signal 22 exceeds a threshold. This essentially amplitude-~orDn~ont pulse-signal on line 25 is coupled to an averaging means 26.
The averaging means 26 pr~duces an ongoing estimate 27 of moisture flow or flux. The circuit ~ , Ls 28-31 comprising the averaging means 26 respond to increases in signal amplitude more rapidly than to decreases. That is, the means 26 features a~y ~LiC attack and decay rates.
This and other prior methods of smoothing the reD~ e of the rain sensor have fallen short of optimum because of underlying principles which will be identified and ~1cc--q~o~ as the present invention is described.
A first limitation of prior art rain responsive systems is that methods used to ~PtormlnD the flow-rate (fluxj of the rainfall are in~o~lu-Le. The prior art approaches typically center around the duration or amplitude of signal excursions of the moi~LuL~ s_nsitive signal. The resulting signal is only loosely correlated to the actual flow rate of the rainfall. Koybayahi teaches (U.S. Patent 4,542,325) that sensor amplitude may be integrated before comparison t:o a threshold, but this does not cir~u~v~llL the effects of small sample size.
Mangler teaches (German Patent DR 40 18 903 C 2~ that the intensity may be derived from the period between s~coccive detections of the sensing means, but this too may be oYrected to vary wildly. The underlying difficulty is that, while they are simple to i ,1~ t, the primarily nmplitude-dopon~ont or period-~eron~Pnt sensing methods of W O 96/37026 ~ 2 ~ 9 3 3 8 ~ PC~rrUS96/06102 -prior art automatic rain re6ponsive wiper control systems do not estimate rain flow-rate (flux) as accurately as does a frequency-oriented method which, as will be described, is implemented in the preferred ~ L of the present invention.
Another reco~ni7ed limitation of prior art rain responsive wiper systems is that, in general, quantities related to the rain flow rate are linear in nature, while most aspects of human perception tend to be logarithmic in nature. That is, amplitude excursions, estimated rate of rainfall, and time-based meaDuL are all linear quantities. While these linear representations are simple to implement in an electronic system, they are at odds with the nature of most aspects of human perception. For example, sound intensity and frequency are both perceived logarithmically by humans, as is light intensity. ~uman perception of time may also be cnn~id~red to be logarithmic in nature. For example, the diff~l~nce between wipe rates of one per second and two per second is considerable, while the difference between ten per second and eleven per second is not readily perceptible. In failing to address the nature of human perception, prior art rain responsive wiper systems inherently L e~ es~11L
large changes in the perceived wipe period with merely small numerical changes in the contents of an internal register within its miuL~col.LL~ller potentially leading to resolution difficulties. It is r~onAhle to suppose that a mi~Lu~L~cessoI based system with a word size of, say, 8 bits is capable of providing a r~Roluti nn more in line with human perception and, hence, can be made to behave in a manner more subjectively pleasing to humans.

P ~ - 2 1 9 3 096~7026 ' 3 81 r~ fS102 A further limitation of known prior art systems is that they operatively contain a single time constant based on prior history of rain a~ tion. The rate of wiper A ctuations should most properly be based upon the 0 5 prevailing, long term (on the order of tens of seconds) conditions in which the vehicle i8 operating, as well as the shorter (on the order of seconds) term fluctuations of t_e sensed rain signal.
The system described in the Koybayahi patent 10 effectively implements a short time constant, the effects of which are ~forgotten~ by the system after each wipe period. The aforereferenced patent 4,620,141 teaches that a time constant may be applied to the sensor output, and this will inherently permit the period of prior wipe 15 ~ lation history to be cnn~i~Pred in dPtPrm;ning an ~p~lu~Llate wipe rate. Similarly, Nangler teaches (German Patent DE 40 18 903 C 2) that the concidoration of prior history can be a function of prior wipe-periods. This also has the effect of ; _l ting a single t;-- _u.lD~anL
20 response. None of the prior art systems r~; nt~ i n~ a separate signal or register value which cuLL~D~unds to the prevailing long-term conditions experienced by the vehicle P~l i rpPd with a rain sensor.

OBJECTS
It is an object of the present invention to overcome the aforementioned limitations of prior art, and thus provide a more subjectively pleasing control of the win~hiPld wiper system.
~ 30 Another object is to provide a means for ~PtPnmin;ng rain flow rate based on the output of the moisture .

~ " 2 1 9338 1 W096137026 -- ' PCT~S96106102 ~

sensitive signal, in a manner which more accurately cGLLe~unds to actual flow rate.
Still another object of the invention is to maintain an internal le~Lese"Lation of rain intensity which more closely corresponds to human perception.
Yet another object is to provide an automatic rain-responsive wiper control system that maintains a long-term average of prevailing conditions, and adjusts the final determination of wiper actuation rate accordingly.
A further object is to provide a means for rapid response to changes in driver-requested sensitivity. This provision is Pcpeci~lly helpful because of the long time constants required to sense prevailing conditions.
A still further object is to provide a control strategy for a rain responsive wiper system which features a propensity for continued operation of the wipers in a given mode.

SUMNARY OF THE lN V
In accordance with the present invention, there is provided a rain responsive win~Rhield wiper control system for controlling the operation of a win~chi~ld wiper motor as a function of the intensity of the moisture striking the windchield. A sensor is dicpo5~d at a pre~etPrmin~
location on the vehicle's wind~hield so that it is traversed by the wiper. The sensor generates an output indicative of the moisture i i ngi ng on the w;n~ch1eld at the location which it occupies. Either analog circuitry or a pL~L -' microcontroller is responsive to the rain event signals from the sensor and repeatedly determin~C
the frequency of uceuLLe.1ce of the rain event signals within a predet~rmin~d time interval to thereby yield a 096mo26 '~ r-l02 recent rain flux value. A unique smoothing algorithm is ~ in the system for effectively eliminating momentary variations in the recent rain flux value. The system also include5 a wiper mode control means that is operatively coupled to the smoothing means for estAhl;~h;ng a sweep speed for the w;n~ch;Pld wiper and it i6 the output of the wiper mode control means that causes the wiper motor to operate at a sweep speed estAhl; ~h~d by the wiper mode control means.
The smoothing algorithm preferably includes a logarithmic conversion feature for converting a recent rain flux value to a recent rain intensity value, it being recognized that human perception in many instances tends to be logarithmic in nature. The smoothing algorithm also ;nnl~ a means for forming a weighted average of medium term rain intensity and long term rain intensity using a multiple time constant averaging approach, the result being that excessively rapid ~ c~ of the wiper actuations to ~hAng;ng conditions is m;n;m;79~.
~K1~L10N OF THE ~~~
The foregoing features and objects of the invention will become apparent to those skilled in the art from the following detailed description of a preferred ~
of the invention, P~pe~Ally when c~nc;d~red in conjunction with the Al - ying drawings in which like numerals in the several views refer to CULL~ ""1;ng parts.
Figure 1 is an illustration showing a rain sensor mounted on a windshield of a vehicle in accordance with the prior art;

W096~7026 2 ~ 9 3 3 8 ~ PCT~596/06102 ~
, ~, ~, .;

Figure 2 is a block diagram illustration of a typical prior art rain responsive windshield wiper control system:
Figure 3 is an illustration of a prior art smoothing algorithm used in implementing the system shown in Figure 2;
Figure 4 is a block diagram representation of the system of the present invention showing the 1 Luved smoothing algorithm of the present invention ~nrlosed in a dashed line box:
Figures 5A through 5E comprise waveforms all drawn to the same time scale and helpful in understanding the operation of the system of the present invention;
Figure 6 are curves illustrating the response of the system of the present invention to a step function input;
Figure 7 is a plot of wiper dwell time vs. rain intensity for the system of the present invention;
Figure 8A is a plot of the rain sensor output vs.
time;
Figure 8s is a plot of the rain deviation signal vs.
time;
Figure 8C is a plot of a hypothetical noise signal V8. time;
Figure 8D is a plot of the overall weighted average of rain intensity vs. time illustrating the effect of hysteresis in mode selection;
Figure g is a software flow diagram depicting rain event detection;
Figure 10 is a software flow diagram relating to disfn~h~nre processing;
Figure 11 is a software flow diagram for the event summation, buffering and intensity conversion employed in the implementation of the present invention;

~ 096~7026 ~ P~ ~ 2 1 9338 I F~l/~6~102 Figure 12 is a software flow diagram of the steps for computing medium and long range averages;
Figure 13 is a software flow diagram for rain intensity and dwell time calculations;
Figure 14 is a software flow diagram for the ba~ky-~u,-d wiper mode control when the wiper system is in its automatic mode;
Figure 15 is a further software flow diagram for system operation in the steady-slow and steady-fast modes;
and Figure 16 is a software flow diagram for implementing instant sensitivity response.

DESCRIPTION OF THE ~h~ V ~MR~DTMFNT
The rain responsive wiper control system of the present invention is shown schematically in block diagram form in Figure 4. This block diagram is i nt~n~ to replace the blocks A/D 11, smoothing method algorithm 12, and vehicle interface 14 in the system diagram of Figure 2. The following text will describe the nature of each of the functional blocks. Detailed flow charts, which one skilled in the art of computer p~uyr ;ng could utilize to write code for implementing the invention, follow after the general descriptions.
Typi¢~l R~in Evnnts ~nd Th~ir Det~,ti-In the present invention, the output from themoisture sensing circuit 1 comprises signal on line 40 which is coupled to an Analog-to-Digital converter 11.
~ 30 The digital output on line 41 from the A/D converter 11 is coupled to an event detection block 42. This block 42 also accepts a digital distnrh~n~e sensing signal on line W096~7026 ~ ~ ' 2 1 9 3 3 8 1 PCT~Sg6/06102 207, which is proportional to a sensed dis~llrh~nre signal from sensor 1 applied to line 43, and is converted by the A/D converter 11 to a digital quantity. SllhsoquPnt processing of the signals on lines 41, 207 are preferably S performed by a p~UyL -' mi~Lu~LucessoI, but may also be implemented in either analog or digital discrete logic circuitry or a combination of the two.
Figure 5A shows typical signals generated by moisture sensing means 1 (after conversion to digital values). The illustration is representative of the sort of signals that might be generated at the onset of a rainstorm, as the rate of rainfall impinging upon the sensor 1 increases from zero to some value over the course of the graph. It should be noted that the preferred sensing device employed may produce either positive or negative signal excursions as the result of an impinging raindrop. Figure 5A also shows that the sensed moisture signal becomes progressively more active over a span of, say twenty seconds. The time scale depicted in this plot may typically be approximately twenty seconds.
The event detection block 42 (Figure 4), described in detail later in the flow charts of Figures 9 and 10, essentially counts the reversals of the rain event pulses 60 (Figure 5A) towards a quiescent value 61 of the digitized rain signal developed on line 41 in Figure 4.
Each such reversal 60 is conci~Pred to be a ~rain event.
Signals in the presence of noise, as at 314, as well as small perturbations, as at 302, are not c~ncidPred ~rain eventsn. A central feature of smoothing algorithm of the present invention is that rainfall flow-rate mea-are taken to be primarily a "number of events per unit time,~ i.e., a frequency ph~~ ---. The driver of the ~ 09~37026 ' 21 9338 I r~ sr ~102 vehicle perceives the flow rate on the w;n~chi~ld primarily as a frequency of impinging raindrops. Thus, the smoothing algorithm in accordance with the present invention is intended to mimic human perception.
5~ It has been observed from experimental data that a consistent flow-rate of rainfall will tend to yield a r~cnn~hly consistent frequency of rain events detected by the sensor. In col.LL~L, the amplitude of successive events varies cnng~rably, and the period between events varies wildly. Thus the primarily amplitude and period oriented schemes of the prior art do not measure the flow rate as accurately as does the frequency-oriented approach used in the algorithm of this in~ention.
While the smoothing algorithm ~igcllccpd herein i8 primarily cnnr~rn~d with the LL~q~"~y of the rain events, as depicted by the pulses 60 in Figure 5A, it is r~ogni7~d that the driver of the vehicle will, in some measure, conciflDr the size of the raindrops as well.
Further, there is some weak c~LL~ nre between size of the impinging raindrop and size of the excursion of the digitized sensed raln signal developed on line 41. Thus each signal excursion 62 in excess of some threshold 303 (shown both positive and negative with respect to ~liescPnt level 61) is counted by~ the ~tec~ i nn block 42 as two events. Similarly, excursions smaller than some threshold 304, as at 302, are not counted as rain events.
Also, when the digitized rnin signal on line 41 goes into saturation, i.e., the digitized rain signal is at a minimum level 65 or maximum leve] 66, the system counts an event for each unit of time (say, 0.3 seconds) during which the signal remains in saturation. As illuDL~Led at - the time identified by numeral 300, this is equivalent to ,~. 2193381 W096l37026 i. ~ C~102 ~qg--ming the OC~ULL~llCe of events when the digitized rain signal has become too large to discern such individual svents. These amplitude-dependent features permit some consideration of drop size, while retAin;ng a primarily freguency-oriented approach.
Note further from Figure 5A that some rain events, like at 301, may occur before the sensed rain signal has L~Ur ,.ed to its quiescent level 61. These signals are also caused by impinging raindrops, and are thus counted as rain events. The method used by event-detection block 42 (Figure 4) for extracting this information will be set forth hereinbelow when the flow charts of Figures 9 and lO
are described.

~ v~.~ion of Events to Event F~
The smoothing algorithm of the present invention det~rmin~q the flow rate of the rainfall by counting the number of rain events in a given interval of time. To do this, the digital values on line 44, ~e~Les~..Ling the number of events, are shifted into a First-In, First-Out (FIFO) Shift Register 45. This register i8 shifted regularly (say, every 0.3 seconds), and events shi~ted out the highest order stage (SFIFO(4)) are forgotten. The total number of events counted within this shift register at any one time are added together by a summing means 46, and the result of this summation is an estimate of the rain flux value or recent rain value (RC_RN_VAL) appearing at the output 47 of the summer 46. A typical signal waveform 305 is illustrated for this flux value 47 in Figure 5B.
It may be seen in Figure 5B that the flux value signal RC_RN_VAL in~L~ ~s upon detec~inn of each rain ~096137026 l3 Ic./~ -c--event. The signal later decrements at a fixed time (1.5 seconds) after the original event, as that event is _hifted out of the shift register 45. For illustration, one such originating event signal 312 and de~L. L 310 is 5pPr;f~r~lly illustrated. By way of this ongoing in~L~ L and de~L~ L to the flux value signal at output 47, an ongoing A~t~rm;n~tion of the number of events per-unit time is maintained. Experiments reveal that this event-freguency method of A;Rc~rning flow-rate (flux) is c~nciderably more accurate than the purely amplitude ~pPnA~nt schemes of the prior art. Additionally, the flux-sensing method inherently provides some smoothing of the digitized sensed-rain signal on line 41, as the number of events per unit time can neither increase to a large level nor decrease to zero instant~n~oucly.
Referring again to Flgure 4, the event detection and totalizing means 42 has provisions for ignoring events in the presence of severe dist~rh~nr~c. Such an event (e.g.
event 314, Figure 5A) is typically the result of a large ambient light dist~lrh~nre le~Lese-.Led by impulse 306 in the plot of Figure 5C. Ambient light distllrhAnc~c, such as might be cause by the shadows of telephone poles and other roadside obaLL~Lions, in some measure effect the sensed rain signal 41. It is desirable that these dis~nrh~nr~c not be c~nc;~red as ~rain events', and a method for rejecting them is described in detail later with the aid of Figure 10.
As previously noted, in the preferred '~ L of the moisture sensing device 1, the resulting sensed moisture signal is bipolar in nature. The terhni~ of intensity meaa~L~ -L set forth in the invention is equally applicable to nn;rol~r signals resulting from W O 96/37026 , ~ ~ q 3 3 ~ ~ P(~rAUS96/06102 ~
.. ~

other rain sensing means, such as conductive, capacitive, or piezoelectric means, as set forth in prior art syst~ms.
For proper discrimination of rain events, it may be n~rr~Ary to place a differentiating means before the event detection block 42 for use with certain of those sensor.

r ~P ~ation and Intensity The RC_RN_VAL output of the rain event detection means 42 and moving summation 46 on line 47 may be considered to be proportional to flux, but it does not yet take into account the effects of human perception. Thus, the algorithm of the present invention couples the ongoing rain flux estimate to a value to intensity conversion means 48 for converting this flux value to an intensity value, RC_RN_INT, at output 49. This conversion means features a logarithmic characteristic, as illustrated by curves 51 - 55 within the fnnrtirnAl block 48. This conversion is described in greater detail by the flow chart of Figure 11. The output at 49 of this functional block 48 may be cnncidrred as an estimate of short-term rain intensity. In this case, uintensity~ implies that the representation is in keeping with human perception.
Cnntlnlling the example set forth in Figures 5A-5C, Figures 5D-5E illustrates typical behavior of the intensity conversion block 48. The input to the intensity conversion block 48 is the ongoing recent estimate of flux, RC_RN VAL. The result of the conversion is illustrated in Figure 5B and comprises signal at 49 in Figure 4. The term, Urecent rain intensity,~ (RC_RN_INT) implies that the signal has cnn~id~red the effects of 096l37026 ~ .t l~ 1 9 3 3 8 1 r~ c~lo2 human perception, but has not yet had the benefits of the multiple time-constant averaging yet to be described.
As mentioned above, a logarithmic characteristic matches the nature of human perception. For example, c~nSi~r the effect of raindrops falling at random locations across the entire w;n~chi~ld at a rate of one per second. This would be perceived by the driver to be a low flow rate, dictating a delay of several seconds between wiper actuations. Consid~r next that adding an additional raindrop each second to the rainfall would double the flow rate. This would be quite noticPRhle to the driver. The driver would desire a decrease in the delay between wiper actuations in order to ~-i ntA i n the same degree of visibility. In contract, consider ten drops per second falling over the surface of the w1n~shi~1d. The flow rate would likely require that the wipers be actuated at the steady slow speed in order to maintain good visibility. Now, the addition of a single drop per second to this condition would be much less noticeable to the driver, were it perceptible at all. The value-to-intensity conversion block 48 fllnrtl nns to mimic this behavior. Note that for larger values of rain flux, the steps between intensity levels become closer together.
A further benefit of incuL~uL~ting this intensity conversion function into the smoothing algorithm of the present invention is that it permits the r~solnti~n required of each of the variables to be evenly spaced.
Say, for example, that an eight-bit variable were used to represent the short term rain intensity at 49. This variable could assume 256 possible levels. Now suppose a change in the flow rate upon the sensor causes a shift in level from, say, 5 to 10. A driver viewing the impinging W096~7026 ~ 2 1 9 3 3 8 1 ~ 6'C6102 ~

raindrops might say that there was a small, but perceptible shift in the rain intensity. Now consider a shift of the same subjective intensity but starting at a higher level, say from 200 to 205. The driver would see this as about the same perceptibility of shift. This even distribution of resolution carries over into many of the other variables and constants which will be ~;ccl~cced~
such as the long range average variable, LRA, at the output 70 of average 59. Because of this even distribution of resolution, 256 possible levels of rain intensity is more than adequate, and the system can be implemented using an ;neyr~ncive~ eight-bit mi.:Lu~locessu~.
In contrast, cnnci~r a smoothing method which lacked this intensity ~ -ncation. In such a case, a shift in a variable Ic~L~s~--ting flow rate from 5 to 10 would be perceived by the driver to be quite dramatic; it is a doubling of flow rate. The possible levels of rain flow rate between these values would be too coarsely spaced for smooth operation of the system. On the other hand, a shift in linearly represented flow rate from 200 to 205 would be imperceptible. Thus the resolution permitted at high flow rates is excessive, and at low flow rates it is inadequate. The same r~Acnn;ng extends to most other variables and constants in the system. The resulting system without intensity -nC~tion would, therefore, suffer performance A~ C resulting from qu~nt;~tion.
Alternatively, more bits o~ resolution could be used to implement the required variables and Cu..~t~l.ts, but this would necessitate a more expensive miuLuULucessuL.
It has been found that different drivers have widely divergent opinions about the desired wiper actuation rate 2~93381 .. ~ ~
096l37026 ' ~ r~ o2 for a given set of conditions. The relative le~u..se of the system to these conditions may be broadly referred to as the system's ~sensitivity," and the driver preferences are - icated to the moi~LuL~ sensitive wiper control system by means of the sensitivity control located on driver accessible switch 15 (Figure 1). The sensitivity sQtting is read into a register in the mi~ucu..LLuller by way of the vehicle interface 14. The sensitivity setting affects the flux-to-intensity conversion, resulting in a family of curves as at 51-55 in Figure 48. Thus, a driver who perceives a given rainfall as being more substantial would set the sensitivity control to a higher level. This results in a higher internal representation of rain intensity.
~ultiple Time-~ - L~L Avrr_ ry As previously mentioned, excessively rapid response of the wiper actuations to ~h~ng; ng conditions is subjectively undesirable. Thus, the output of the short-term rain intensity estimate at 49 is coupled to a mediumrange averager 56 for taking a medium range average of the signal. The medium range average is ;mrl~ ed with different time constants for attzck and decay. For example, attack time may be 1.2 seconds and the decay time is 5.5 seconds. Such an ~v~agel can be constructed using a resistor/diode/capacitor network 57, as shown schematically in Figure 4, or alternatively and preferably, this can be realized in software (Figure 12).
Referring to the plot of Figure 5E, a typical response for the medium range average is labeled as such.
It can be seen from this illustration that the medium - range average signal at output 58 increases, or attacks, W0 96/37026 2 1 9 3 ~ 8 1 F~ 6C102 -more rapidly than it decreases. See curve s~_ Ls 320 and 322. The medium range offers only a modest amount of smoothing over the recent intensity signal, permitting re~con~hly rapid response.
In addition to maintaining a medium range average, the smoothing algorithm of the present invention ~Lu~oses that the output 58 of the medium range averager be coupled to along range averager 59. Thus, the already smoothed rain intensity signal is further smoothed over a longer period of time as detPrminPd by the long range av~L~geI.
For example, the attack time constant for the long range averager may be two seconds and the decay time cull~L~-lL
ten seconds. The resulting signal at output 70 le~L~SeI~LS
the long range prevailing conditions under which the vehicle i5 operating. This, again, is intended to mimic human perception. When the driver of a vehicle Pq~ipped with manually adjustable wipers sets the wiper speed, he or she will cnncider the prevailing operating conditions, as well as the current intensity of rainfall.
Also illustrated in Figure 5E is a curve showing the typical result of this long range averaging by average means 59 ~ As the signal at output 70 is intpn~pd to indicate the long-range prevailing conditions, it increases and decreases more slowly than does the medium range av~L~gel-.

Overall ~e~ghted Average The outputs 58 and 70 from the medium range ~veL~
56 and the long range dV~ l 59 are applied to a summing means 71 which performs a weighted summation of the medium and long range average signals. The output 72 of this summation means 71 may be cnnci~pred to be the overall I c~ 1 9338 1 096~7026 PCT~s96/06102 rain intensity estimate. The system so described thus produces an estimate of the rain intensity which is proportional to the human-perceived rate of rainfall impinging upon the windshield. Because of the effects of the separately maintained long-range average, the resulting rain intensity estimate cnncid~rs both the recent history of the rain sensing signal (at 58), as well as the prevailing long-term conditions under which the sensor is operating (at 70). A typical overall weighted average signal is so labeled in Fiqure 5E.
Because of the cnnci~ration of long-term conditions, the system of the present invention Le~ollds very rapidly to changing conditions, but the permitted range of response is limited, more so than it would be if only a single time constant is utilized in the averaging means.
The system tends to keep the le~..se in the range of the value maintained in the long range ~Vel~geL 59. This makes the system of this invention smoother than a single t;r- cu..~Lant system could be (due to the long range averaging), yet faster to respond to a rapid, small changes in conditions.
The response of the system to a step function change in rain intensity is illustrated in Figure 6. At time t the system encounters the step function stimulus which warrants operation of the wipers in a steady slow fashion.
The medium range averager 59 will respond very quickly, so the system will very quickly achieve a fast intermittent speed (a rate above the level i~Fntified by numeral 83).
The long range averager 59 will respond more slowly, and the system will then shift into a steady slow mode of operation. There is in effect a ~knee~, as at 82, in the time-domain response. It may also be said that the system ~ 1 9 3 3 8 1 W O 96137026 P(~rrUS96/06102 very quickly responds to a level which, while not optimal, is not ob;ectionable. The system then smoothly ~fine tunes~ itself to the optimal level.
For purposes of illustration, the les~unse of a typical prior art system is superimposed on Figure 6 as curve 84. Without the benefit of the response knee 82, the resulting single time constant will be subjectively too sluggish or too erratic. Quantitative experiments reveal that the present system is, in fact, 66% smoother than prior art systems, with smoothness being taken as the ratio of standard deviation to mean wipe period for a large number of wipes. Also, the present system is simultz-nPol~c~ y 20% more rapid to respond to step function changes in rainfall flow rate, where the response time is --- ed to 90% of settled, as is customary in the electronics industry.

Mode of Oper~tion The actual mode of operation of the wipers is based on the rain intensity (RN_INT) estimate at the output of sunming means 71 of Figure 4. This is coupled to a functional block 73 that, by way of the vehicle interface 14, runs the windshield wipers 4 at a rate a~ Liate for the conditions. For small values of rain estimate, the wiper6 are run in an intermittent mode, with long delay times between actuations. If the rain intensity is higher, the delay time between wipes is made shorter. If the rain intensity should exceed a threshold, the wipers will ~ a operation at a steady slow speed, with no delay between successive wipes. If the rain intensity exceeds a value higher still, the wipers will operate at a ~ ~096/37026 i~ & ~ ~ 2 1 9338 ~ PCT~S96106102 higher rate of speed. To further prevent erratic behavior, there is c~nci~Prable hysteresis among the 510w and high speed thresholds.
The characteristics of the mode determination are illustrated as a graph in Figure 7, and flow charts ~t~; 1 i ng its microprocessor implementation are presented in Figures 14 and 15. The effects of the hysteresis between the fast and 810w speeds is illustrated for a typical signal in Figure 8D. (Note that this signal does not coLLe~ond to the same signal set forth in Figurefi 5A-5F, or that of Figures 8A-8C.) In figure 8D, the vertical axis of the graph is the overall weighted average of rain intensity at 72 in Figure 4, as computed previously, and the horizontal scale spans about twenty seconds of time.
As can be seen, the initial value of rain intensity between to and t~ is above the threshold 226 for operation in a slow, intermittent mode, and below the threshold 229 for a transition from slow to fast mode of operation.
Thus the vehicle interface will operate the wiper motor at a steady slow speed. Although the rain intensity signal at 72 fluctuates conci~prablyl it does not cross the threshold 228 for a transition from slow to fast speed until time t3, bheleu~on the wiper motor 13 will ~ -e operation at a steady fast speed of operation. From this point forward, the rain intensity signal at 72 remains above the threshold 229 for transition from fast to filow operation. Thus the wiper motor 13 will continue to operate at a fast speed, despite fluctuations below the slow-to-fast threshold 228.
It is clear from examination of Figure 8D that were there but a single threshold (for example, at 9l) to determine operation in slow or fast speeds, without the W096~7026 ~ 2 ~ 9 3 3 8 1 PCT~S96/06102 benefit of the illustrated hysteresis, the illustrated signal would have resulted in several additional transitions between slow and fast speeds. This would be ~udged by the driver to be erratic operation. The present invention may be thought of as maintaining a propensity to operate at a single speed.
Values of rain intensity below the intermittent/slow threshold will naturally require that the mode ~t~rmin~tion block 73 incoL~ul~te some means for computing an appropriate dwell time. The methods for actually implementing such a dwell time are inherently linear in nature. A mi~LupLuce~_uL may, for example, be pluyl -1 to implement a delay of some variable number of seconds. This linear implementation of time is incongruent with normal human perception, in a manner similar to the example of human perception of flux previously described.
Thus, in order to A ~ te normal human perception, and to _ --te for the logarithmic nature introduced in the value-to-intensity conversion block 48, the present invention employs a means in block 73 for converting logarithmic intensity to liner time. This behavior is illustrated in the curves of Figure 7. The input to block 73 is the overall weighted average of rain intensity at the output of summing means 71, and is plotted on the horizontal axis in Figure 7. The output of this functional block is exponential in nature with decreasing input. A software means for implementing this behavior is set forth in the flow chart of Figure 13.
Also, in this invention, the computed wipe period is multiplied by a constant which changes with the ? r 2 ~ 9338 ~
ONO 9613~026 r~ 102 sensitivity input provided by the driver. This feature is illustrated in the family of curves 86 - 88 set forth in Figure 7, and further ~ d~tes driver preferences.
Smooth operation in the transition between intermittent and steady slow speeds is also a consequence of this control strategy. Observe that if the rain intensIty signal illustrated in Figure 8D were to drop ~ust below the intermittent/slow threshold 226, the resulting computed dwell time would be zero or some small value. A transition from steady slow to, say, one half second of dwell between wipes, i8 not objectionable to the driver. Thus, the driver will perceive the operation of the system as smooth, even with modest fluctuation of the rain intensity signal;
Any rain sensor 1 mounted within the path swept by the wipers 4 must deal with the effects of the wipers passing over the sensor. In the present invention, if the dwell period between wipes is long (say three or more seconds) it is proposed that data from the rain sensor be ignored while an actuation of the wipers is in ~L~yL~58.
Further, in such a situation the currently r-in~;n~A
short, medium and long range ~veL-y~S of rain intensity should be held constant for that time.
For shorter computed dwell times, this method would prove undesirable in that it would ignore a large percentage of the data coming from the rain sensor. Thus, the present invention features a means of ignoring a certain number of rain-detection events after the start of an actuation of the wi n~ch ~ ~ d wipers. This roughly - 30 ~, tes for the effects of the wet wipers traversing the sensor. Should the wiper blades be dry, thus not generating a large signal when traversing the sensor, the C~ ~ 2193381 W096137026 PCT~S96/06102 proposed method will overcnmr~nCAte, causing the system to ~orget legitimate rain events. This is desirable, because the fact that the wipers are dry, and thus prone to smearing, is valuable information that should not be discarded. Further, the invention requires no synchronization to the exact moment the wipers pass over the sensor.

Instant ~ "~ to 6ensitivity Changes It is desirable that the system respond instantly to changes in the sensitivity level of the system estAhli~hnd~
by the vehicle's driver. Thus, should the driver increase the sensitivity, the ongoing medium and long range averages applied to summing means 7l are immediately increased. Conversely, if the driver decreases the system sensitivity, the system will instantly decrease these averages. This feature has the effect of immediately ~hAnqi ng the behavior of the system upon reception of a new desired sensitivity. The driver need not wait for the time constants in the averaging means to reach a new level. A method for the software implementation of this feature is set forth in the flow chart of Figure 16.
The above description generally illustrates the desired behavior of the functional blocks which make up the wiper control system of the present invention.
Figures 9 - 15 comprise a number of flow-charts, which describe a detailed description of the invention based upon a pLU~L ~ miulucu~l~Luller implementation. One skilled in the art of ~OyL ing will see that the A'_, ylng flow charts teach a detailed method of i~rl~ Ling the smoothing algorithm set forth in the block diagram of Figure 4. The symbols of the flow chart 096~7026 ' ~ ~~ ¦ A PCT~S96106l02 may be implemented in some direct machine-executable language, such as asse~bly code, or with the aid of some higher-level compiled language, ~uch as UC.~ For convenience, set forth below in Table I is a listing of variables used in the flow charts and Table II ldentifies constants identified in these same flow charts. Some of these variables COI r ~ol-d directly to digital signals illustrated on the smoothing algorithm overview block diagram of Figure 4. Variable and constant names appear in upper case.

Table I:variables used in the A' _ ,-nying flow Ref charts No.
RAIN The results of the analog-to- 41 digital conversion of the moisture-sensing signal. (This ~ur~ -ds to the signal referred to as Uthe moisture sensing signal.) RC_RN VAL Recent Rain Value. This is the 47 total nu~ber of events in the First-In, First out. shift register 45.
RC_RN_INT Recent Rain Intensity. 49 2ID MRA Medium-Range Average of the Recent 58 Rain Intensity. (This is a sixteen bit quantity considered to range between zero and one.) :

2 1 933~ ~
W096l37026 r~ 'C'I'~

Table I:variables used in the A~ nying flow Ref charts No.
LRA Long-Range Average of the Recent 70 Rain Intensity. (This is a sixteen bit quantity c~n~;~Pred to range between zero and one.) RAIN_INT Rain Intensity. This is 72 essentially the output of the rain-intensity sensing portion of the algorithm. It is a weighted average of the medium and long-range averages.
DWELL_TIME Dwell Time is the nominal delay Fig. 7 between actuations of the wiper, and is computed based on the RAIN_INT. Given in units of half-seconds.
BLANK_CNT A counter used to time periods 200 wherein rain sensor data is ignored.
5 RINCR Rain Increasing. A flag used in 201 rain event ~Ptecti~n.
RC_XTRM Recent extreme of rain deviation. 202 Used to extract maxima and minima of the rain deviation signal for purposes of event detection.
RAIN_DEV The deviation of the moisture 203 sensing signal from its nominal quiescent value.

~ ~096~7026 ;.~ ~ 2 1 9 3 3 8 1 p~ 6l02 ~ 27 Table I:variables used in the ~ nying flow Ref charts No.
SFIFO[0] Summation First-In, First Out shift 205 register array.

SFIFO[4]
SNS_VAL Sensitivity Value. A number 206 between 0 and 4 inclusive which indicates the setting of the driver-accessible sensitivity control.

NOISE The results of the analog to 207 digital conversion of the disturbance sensing signal.
OLD_SNS_VAL The previous reading of sensitivity 210 value.
DWELL_COUNT Used by the mi~ cessoL to 212 generate delays of DWELL_TIME.
Thus, it in~L~ ~ every one-half second.~0 NOISE_DEV The absolute value of the deviation 214 of the noise signal from its normal quiescent value.

W096~7026 '~ 2 1 9 3 3 8 1 rc~ c-~02~

TABLE II

Table II: Table of constants used in Ref No.
~ nying flow charts:
K_INT Multiplying constant used in 215 value-to-intensity conversion.
Adjusted so that RC_RN_INT has an ~,uLu~Liate range of values to optimally utilize 8 bits of resolution.
K_SHIM Constant added to SNS_VAL in 216 value-to-intensity conversion so that an a~upLiate range of adjustment is permitted to a~ te driver sensitivity preferences.
K_DW_BP Dwell Breakpoint Constant. Above 217 this value of rain intensity, the nominal dwell time should be zero.
Set to approximately the same value as SLW_INT_THR 227.
K_DW_EXP Multiplying CUllaLdllL used in the 218 exponential conversion of rain intensity to dwell time.
K_DW_MUL Multiplying cul-aLal,L used to 219 adjust the result of the conversion to dwell time to a suitable range of values.

0~096f37016 I~ 'C'102 Table II: Table of constants used in Ref No.
nying flow charts:
K_ET Event Threshold. Excursions of 220 the sensed rain signal must be at least this many units to be considered valid ra~in events. Fig. 8 K_ATT_MRA Attack constant for the medium 221 range average.
K_DEC_MRA Decay constant for the medium 222 range average.
K_ATT_LRA Attack constant for the long range 223 average.
K_DEC_LRA Decay .;.na~a~ for the long range 224 average.
OFF_THR Off Threshold. When RAIN_INT is 225 below this value, t.he mode control program enters the Automatic-off mode.
INT_SLW_THR Intermittent to Slow Threshold. 226 Used to determine when the wiper system should transition from intermittent to a steady slow mode Fig. 8D
of operation. The next three thresholds are similar in their definitions.
SLW_INT_THR Slow to Intermittent Threshold 227 SLW_FST_THR Slow to Fast Threshold 228 FST_SLW_THR Fast to Slow Threshold 229 ~ t ; 2193381 W096~7026 F~l/~50l02 Table II: Table of constants used in Ref No.
a -nying flow charts:
K_LDTH Large Drop Threshold. Deviations 230 of the sensed moisture signal which are larger than this threshold are c~n~ red to count as two events.

EVENT n~tTIt As can be seen from block 100, in the flow chart of Figure 9, every 9 mill;c~ron~c the mi~L. _Ler pe~ful~_ a software interrupt, whereupon the steps comprising the method employed for detection of rain events is initiated.
This sampling period COL r ~ nds to a sampling r~ ~U~ y of about 110 Hertz, and it can also be used to generate hy~pLiate infrared emitter timing pulses, as described in the aforereferenced U.S. Patent No. 5,059,877. The generntion of these pulses is ;n~p~n~nt of the smoothing method described here, and, thus, is not illustrated.
Upon execution of the software interrupt (block 100), the microcomputer implements a ~C;cio~ block lOl which permits the bypassing of the rain event detection code.
As illustrated, when the variable BLANK CNT (Table I, Ref.
200) is nonzero, rain events are ignored. This variable may be set in the overall mode control program (Figures 14 and 15) during the execution of a single actuation of the wipers (block 149). The purpose of this implemented feature is to prevent the retriggering effect of the w;n~chi~ld wipers sweeping water over the rain sensor leading to an excessively high estimate of the rain intensity.

' i 2 ~ 9 3 3 8 1 ~096~7026 P~ E~C102 At block 102, the miuLuc _Ler has implemented the software interrupt and de~Prm;nPd that the data from the sensor is not to be ignored, and the A/D converter reads the moisture sensing signal into the a~ Liate variable RAIN at 41 in Figure 4. In the following block 103, the deviatlon (RAIN_DEV; Table I, Ref. 203) of the sensed rain signal from its nominal ~l;Pccpnt value 61 (Figure 5A) is extracted, using an absolute-value function. A typical digitized rain signal and resulting deviation (RAIN_DEV) are illustrated in Figures 8A and 8B, in order to illustrate the process of event detection.
At this point, a decision block 104 implements different code, based on whether the RAIN_DEV signal was previously increasing or decreasing. This information is maintained in a bit flag, RINCR ~Table I, Ref. 201). If the rain deviation is increasing, the operations e~essed in blocks 105 and 106 are executed to search for a new recent extreme maximum value of the rain deviation signal, RC_XTRM (Table I, Ref. 202). Thus implemented, RC XTR~
peak detects the rain deviation signal. The next ~r; c; nn block 107 looks for a reversal o~ the rain deviation signal towards zero. It does this by ascertaining that the rain deviation signal (RAIN DEV) is less than the recent extreme value (RC XTR~) by at least the value of a threshold, K ET (Table II, Ref. 7.20). It is this reversal from a peak excursion of the deviated rain signal which we define as a Lrain event.~ Upon detection of this rain event, the function performed at block 108 inuL~ c the first stage of a summing first-in, first-out shift register 45 (Figure 4), SFIF0[0].
The next executed decision indicated by block 109 compares the deviation of the rain signal 203 (RAIN DEV) ~ ! 2 1 9 3 3 8 1 W096/37026 ~ PCT~S96106102 against a threshold, K_LDTH (Table II, Ref. 230).
Excursions in excess of this threshold may be considered to have originated as the result of larger raindrops, and are thus counted a second time into SFIF0 (0). See .
functional block 110. Thus, the operations le~s~ ed by blocks 109 and 110 have the effect of weighing the eDp~..se to the rain events by the size of the rain signal excursion. This represents something of a ~ ~ ice with the purely amplitude ~pen~nt schemes of prior art approaches. Additional sensed levels may be i 1 Led by o~cc~ing sets of blocks similar to 109 and 110, each set implementing a different threshold. It has been found that four such sets, only one of which is illustrated, is sufficient.
Next, the RINCR flag (Table I, Ref. 201) is reset to zero (block 111). This signals the program so that on the next pass through, the RAIN_DEV signal is considered to be decreasing. Program control now passes on to the distllrh~n~e proc~cci ng block 112 in Figure 10.
With continued reference to Figure 9, as mentioned above, in the case where the recent RAIN_DEV signal has been decreasing is handled differently at decision block 104. In such a case, new valleys in the recent rain deviation variable are detected by way of d~cicion block 113 and function block 114. These valleys are stored as the variable RC_XTRM (Table I, Ref. 202), which in this context now meas an extreme valley, rather than an extreme peak. Next, a d~cicinn indicated by block 115 detects if the rain deviation variable has begun increasing towards some new peak again, by at least the amount of the event threshold, R_ET (Table II, Ref. 220). If this is the case, at operation block 116 the flag RINCR is set to 2 1 9338 ~
~VO96/37026 E~ 1/U.. 6~ 02 indicate that the rain deviation is now increasing. It can be said that the system is one again armed to detect another rain event. In all cases the program proceeds with distnrh~nce processing 112 ~Figure 10).
Referring to Figure 8A, the process for detecting an event may be illustrated as follows. At the start of the graph the rain deviation is zero, and the flag RINCR i8 reset. The time span of the illustration is taken to be about two seconds, so the software interrupt (block 100 of Figure 9) occurs hundreds of times over the course of the graph. At the beginning of the graph in Figure 8B, at each pass through the event detection flow chart (Figure 9) ~'~cicion block 104 selects the ~yes~ branch which will sllhsPq~lQntly look for a condition where the rain deviation lB is increasing (block 105~. h'hen a drop of water impinges upon the sensor (as at time 92 in Figure 8B), the digitized rain signal tRAIN~ begins to deviate from the q-iecc~nt level 61. This results in a rain deviation (RAIN_DEV) signal which inuL_ases by more than the threshold K_ET (Rain event threshold constant), ~h~L~U~UII
block 116 will set the flag RINCR, arming th system for a reversal detection.
The RAIN_DEV signal continues to rise until it peaks and p~ùceeds back towards zero. Once it has decreased an 2!; amount set by K_ET, the event i5 detected and flag RINCR
is reset at block 111. In the illustration of Figures 8A
and 8B, a second drop (at time 93) causes an excursion of the rain deviation signal which is at least a threshold value K_ET qreater than the lowest valley 94 in the signal 30 (held by RC_XTR~). Thus, this is also detected as an event. Note that it is not n~C~CC~rY for the rain deviation signal to drop below a fixed threshold in order ~i ' : 21 9338 1 W096~7026 ~ 'C102 to detect a new event; the method illustrated detects reversals.

DI~I~AUCE ~ ~ e As shown in the flow chart of Figure 10, and continuing the g mi 11 i secnn~ software interrupt, block 117 utilizes the A/D converter 11 (Figure 4) to read the value of the distl~rhAnre sensing signal into a variable ~noise deviation" ~NOISE" (Table I, Ref. 207). The operation called for at block 118 then computes the absolute value of the deviation of this variable from its nominal value, leaving the result in the variable ~noise deviation~
NOISE DEV (Table I, Ref. 214). This variable i8 then compared against the rain deviation variable. (See block 119.) If the magnitude of the noise deviation exceeds that of the sensed rain deviation, the authenticity of the rain event in progress is suspect. That is, it is plausible that a sharp shadow or supply voltage fluctuation initiated the current deviation of the rain signal in progress. Thus, in such case, the operation indicated by block 120 resets the flag RINCR to zero.
This has the effect of discounting the rain deviation in progress.
Next, and in all cases if the rain deviations exceed the distnrh~nre deviations, the program returns from the 9 ml 11 i~ecnnd software interrupt (block 121).
The effects of a typical distl~rh~nre event is also illustrated in Figures 8A and 8C. In the graphs, at a time indicated by numeral 95, a sharp shadow sweeping across the sensor 1 causes a change in the digitized rain sensing signal (RAIN). A short time later (at 98 in Figure 8C), this same di~l rh~n~e effects the noise ~'096137026 P~ , S.lC~I.

sensing signal (NOISE). Note that in this example, selected to illustrate a particularly troublesome case, distnrh~nre event at time 95 actually modulates the rain signal on line 41 just before it affects the noise sensing signal (NOISE). Because the distllrh~nre signal deviation (NOISE_DEV) exceeds the rain signal deviation (RAIN DEV) at time 96 before the reversal of the rain deviation signal at time 97, the RINCR flag is reset, and, thus, no rain event is counted.
If suitable delays (not illustrated) are added in the acqulsition of the variable, RAIN, then all distllrh~n~es close in time to rain events can cause rain events to be ignored, including those that o~uLL~d just before the distllrh~n~e. The terhni~l~ is described in detail in U.S.
Patent 5, OS9, 877 referred to earlier.

__ - AND FIPO
Rain events, when detected as described above, may occur many times in a second, ~PpPn~ing on the conditions and sensed area. Thus, it is nPc~cc~ry to perform the routine many times each second. The 1l in~Pr of the rain intensity estimation method may occur with relative infrequency, as it is not nPcpcc~ry to re-adjust the speed of the wipers more than a few times per second. Thus, the rest of the rain intensity estimating method, as dPpicted in the flow charts of Figures 11 through 13, is pelL- ~' every 0.3 seconds, driven by a software illL~LLu~.
Upon reception of the software interrupt (block 122, Figure 11), the operations indicated in the three functional blocXs 123, 124, 125 serve to total all of the counts in all of the stages of SFIFO 45, leaving the variable RC_RN_VAL (Table I, Ref. 47~. This guantity is } .

I ~c~, ~ 2 1 9338 ~
W096l37026 ~ C-102 here referred to as "recent rain flux value," as opposed to intensity, because no tion of the effects of human perception has yet been factored in.
Next, each stage of SFIFO 45 is shifted to the next higher stage. The lowest stage, SFIFOtO] is reset to zero. This operation is shown in operation block 127 of Figure il, and effectively implements the first-in, first out shift register.
With reference to block 129, the abuv~ ubL~ined rain value, which may be thought of as Uevents per unit of time,~ is converted to an intensity value. The operation represented by functional block 128 imparts a logarithmic characteristic to the value, causing the resulting quantity, RC_RN_INT (Table I, Ref. 49), to match human perception of rainfall intensity. This recent rain intensity value is increased with increased setti ng~ on the driver-~cces~;hle sensitivity control, as depicted in the Figure 2, as part of the driver Arc~ihle switch 15.
This feature, in some measure, ~ tes driver prefeL~nces. That is, one driver might judge a condition to be a 'light rain,~ while another driver may consider the same condition to be 'heavy ~ ,uuUI . ~ The modulation of Recent Rain Intensity with sensitivity changes ~epl s~..Led by block 129 reflects these differing opinions.
MULTIPLE TI~E-CON8TANT AVrD~ I
The output of the above value-to-intensity conversion, RC_RN_INT, thus matches human perception of rainfall intensity, but it fluctuates c~n~ rably due to the restricted sample size available to the detector.
Referring to Figure 12, a medium range averaging operation is performed on the recent rain intensity (RC_RN_INT), as ~ 096/37026 2 1 9 3 3 8 1 r~ s o~lo2 , J~~

depicted in blocks 131 - 134. These operations produce an output, MRA (Medium Range Average) which asymptotically approaches the input value, RC_RN_INT. This characteristic is similar to that provided by a resistor-capacitor filter 57, a,s illustrated in the schematics of Figure 4. As indicated, the implemented filter has a relatively short response to new rainfall, or a short attaclc time, of about 1.2 seconds. The recovery of the filter in the absence of rain, or the decay time, is longer at about 5.5 seconds. Thus, the output of bloc~c 56 in the overview schematic of Figure 4 provides an estimate of the medium-range rainfall conditions, averaged over a preceding several seconds.
As previously ~icc~ese~ no signal combination of a~y LLic attack and decay time cu-.~tallt~ can satisfactorily implement a filter which is ~udged simult~neo~ely smooth and responsive. Thus, the output of the medium range averager 56 (MRA) is transferred to the input of a long range averager 59. The purpose of the long range average, implemented by the operations represented as blocks 135 - 139 in Figure 12, is to estimate the prevailing conditions under which the sensor is operating. In a similar fashion to that of the medium range average, the long range ~v~L~r provides a~y t ic asymptotic attack and decay constants. As this is fed from the output of the already-smoothed medium range averager 56, the resulting time constants are on the order of ten seconds or so. The system may alternatively be implemented with the long range averager 59 in parallel with, rather than in series with, the medium range aveL~geL. This would require longer time constants in the long range averager. It i8 n~Ce~SAry to maintain 16 ~its -W096137026 t '~ 2'~ q338 1 .~ C102 ~

of resolution for both the long and medium range averagers, in order to i~pl~ L the long time cunDL~l-Ls reguired.

RAIN l~ ~l'L'I A~D DWELL TIME
With medium and long range averages of rainfall intensity est~hlich~d, the smoothing algorithm pLuceeds to construct an overall estimate of rainfall intensity, RAIN_INT. This quantity is the average of the medium and long range averages as performed by operation block 141 in the flow chart of Figure 13. The resulting quantity, RAIN_INT, provides the desirable bounded ~u..-e to rapid changes in conditions, while maintaining some consideration of long-term prevailing conditions.
Snhce~lont determinations of wiper activity in automatic modes of operation are based on this quantity.
For resulting rain intensities which lie below a threshold K_DW_BP (Table II, Ref. 217), it will be desirable to operate the w;ndchi~ld wipers with some dwell time between wipes. Specifically, the next step taken by the mi~Lucu..LLuller, as shown starting at d~lcinn block 143 in Figure 13, is to compute the required dwell time.
Because RAIN_INT is proportional to human perception of rain intensity, and this is, in turn, proportional to the logarithm of the actual measured flow rate, it i8 n~C~cc~ry to introduce an ~ ne..Lial characteristic into the ~ ~_L~tion of dwell times. This is the purpose of the operations reflected at block 145. That is, as RAIN_INT falls further below a threshold tRef. 217, Table II), the computed DWELL TIME grows exponentially, as illustrated in the curves of Figure 7. To further te the system sensitivity preferences of the ~'096~7026 ~ 2 ~ 9 3 3 8 1 PCT~S96~6102 driver, this dwell time is multiplied by a con~La..L
R_DW_MUL block 146 which is changed with sensitivity (SNS_VAL) (Table I, Ref. 206). As illustrated in Figure 7, at higher sensitivities, a shorter dwell time is computed for a given value of RAIN_INT.

MODE CONTROL
The flow charts of Figures 14 and 15 illustrate the background wiper mode control of the wiper system in the automatic mode of operation. The~e flow charts assume that the previously described rain estimation algorithm (Figures 9 - 13) has established an estimate of rain intensity and an app~u~liately computed dwell time.
Provision for manual operation of the wipers i5 not illustrated. Such control strategies are well understood and readily implemented by one skilled in the art of ~Gi~n; ng wiper control systems.
Upon entry into automatic mode (block 148), the mi~Lucul.LLulier executes a single actuation of the wiperG
(block 149), and then turns the wiper motor off (block 150). At this point (block 151), the dwell loop counter, DWELL_COUNT is reset. The mi~Luuul-LLuller then pLuceeds to execute the dwell loop shown enclosed by dashed line box 167. The first step in this loop is to compare the Rain Intensity with a threshold INT_SLW_THR (Table II, Ref. 226) at block 152. The threshold, INT_SLW_THR, is set at a level ap~luyLiate for a transition from intermittent into steady slow mode of operation. If RAIN_INT, in fact, exceeds this threshold, the wiper motor control is transferred to the steady slow mode (block 160, Figure 15).

, W096~7026 , ~ 9 3 3 8 1 PCr~S96/0610 The operation reflected by d~iRinn block 153 det~rmin~c if the RAIN_INT variable is low enough to justify entry into the Automatic-off loop which is shown enclosed by dashed line box 168. The purpose of this loop comprising operation block 157 and ~Pc;~ion block 158, is to hold the wiper motor off in the period of a prolonged nbsence of sensed moisture. Decision block 158 transfers control so as to execute a single wipe upon detection of sufficient rain intensity. It is contemplated that the driver may prefer that the system execute a single wipe every several minutes even in the absence of sensed moisture. If the vehicle manufacturer deems this to be the case, blocks 157 and 158 may be modified accordingly.
Continuing with the description of the flow charge of Figure 14, if the Rain Intensity is still within a range a~ Liate for operation in an intermittent mode, A~c;ci~n block 154 ~PtPrm;neC if the computed dwell time tTable I, Ref. 85) has expired. In such a case, the system p1vceeds to execute another single wipe (block 149). As mentioned earlier, the appropriate DWELL_TIME is frequently being L~ __Led by the rain estimation algorithm delineated in the flow charts of Figures 9 - 13.
Thus, a sudden increase in rain intensity can cut short a dwell interval. Also, a sudden increase in the user-detPrm;ned sensitivity will have the same effect.
If the dwell interval has not expired, as det-~rm;ned by the test at block 154, the operations depicted at blocks 155 and 156 will delay one-half second and in~ t the dwell counter varlable, DWELL_CNT. Program control is then transferred to the beginning of the dwell loop 167.

O 96137026 ~ ; ~ . 2 1 9 3 3 8 I PC~rrUS96/06102 Should the aforementioned dec;cio~ block 152 determine that operation in steady-slow mode is a~IupL1ate, program control will proceed, beginning at block 160 of Figure 15. Upon entry into the steady slow mode, the wipers are made to run at a slow speed (block 161). The operation reflected by block 162 will dnt~rm;nn if the RAIN INT variable has decreased to a point (SLW_INT_THR) where steady slow operation is no longer =desirable. If this is the case, control is transferred at block 166 to the Automatic Mode (block 148), whereupon the system will enter the intermittent mode of operation. The thresholds of comparison, SLW INT_THR and INT_SLW_THR, ~nd indeed all thresholds cnnc~rn;ng operational mode of the wipers (Table II, Ref. 225 - 229), are set in such a way as to allow considerable hysteresis with respect to RAIN_INT, in the example A;ccllc~A earlier with the aid of Figure 8D. That is, if the system is operating in a given mode, it will tend to continue to operate in that mode.
With continued reference to Figure 15, ~ec;C;nn block 163 will compare the rain intensity RAIN_INT with a threshold (SLW_FST_THR) to Aet~rm;n~ if the rain intensity i8 sufficient to warrant high-speed operation of the wipers. I~ this is not the case, branch 172 will be followed to continue to keep control of the wipers in a loop at the steady slow speed. Otherwise, the program control will transfer to block 164, which will run the wipers at fast speed. The test at A~r; ci~n block 165 will determine whether to hold control in the steady-fast loop 169 until RAIN_INT has decreased through the threshold FST_SLW_T~R and low enough for operation at steady slow speed. In such case, branch 170 will be followed to transfer control to the steady slow loop 171. Thus, the t, W096~7026 3 8 1 PCT~S96/06102 flow charts of Figures 14 and lS fully describe the control of the wipers in the automatic mode of operation.

~,,.,, I ,,~I.~T. Co21TB.OL
As previously mentioned, control schemes for manual operation are not illustrated here. Other mi~L~L~ce~s~L
based interface functions, such as reading the driver-acc~csihle switch, implementing a wash function, and methods of operating the wiper motor, are also conci~red to be readily implemented by one skilled in the art of automotive wiper controls.
The instant sensitivity response, as previously mentioned, is a desirable feature of this invention, and Figure 16 depicts a flow chart for its implementation. The operation of block 176, which is specific to the vehicle interface, reads the position of the sensitivity control into a register SNS_VAL (Table I, Ref. 206). This occurs in a background control loop, and is executed several times every second. As indicated by d~cicisn block 178, a comparison is next made with the previously read value of sensitivity, OLD SNS_VAL. If the sensitivity has increased, branch 190 executes, and the operation at block 188 increases the medium and long range averages by a factor of 1.5. This has the effect of reinterpreting the history of the stimulus, and the previously described mode control will instantly respond to the change.
Alternatively, if at decision block 180, it det~rmin~d that the sensitivity value has just decreased, medium and long range averages are both multiplied by a factor of 0.66, at block 182 and mode control will instantly slow down operation of the wipers. In either case, the operation at block 184 will update the value of 096~7026 P~ ~ 'C1~2 OLD_SNS_VAL (Table I, Ref. 210), and the process;ng of the instant sensitivity response snhsD~Dntly terminates as indicated by block 186.

~ .el~ion All of the features described in this invention serve to address an intractable problem, i.e., how to devise a control strategy that functions a~ u~ iately with signals from a sampled area which is very small in comparison to the size of the entire windch;Dld. Further, they address the conflicting requirements that the resulting system must be simultaneously smooth and responsive. The solutions employed seek to mimic human perception. The net result of all of these features is a system which in field evaluations has proven to be subjectively very pleasing to the driver. Thus, the driver of the vehicle typically soon learns to forget about the operation of the win~chipld wipers, which is the ultimate objective of a moisture-sensing wiper control system.
This invention has been described herein in c~nci~Drable detail in order to comply with the Patent Statutes and to provide those skilled in the art with the information needed to apply the novel principles and to construct and use such spDci~li7ed ~ ~5 as are required. However, it ls to be understood that the invention can be carried out by specifically different ~ nt and devices, and that various modifications, both as to the ~ details and operating P~CedULe8 can be accomplished without departing from the scope of the invention itself.

Claims (20)

44 What is claimed is:

1. A rain responsive windshield wiper control system for a vehicle of the type having a windshield, a windshield wiper adapted to be driven by a windshield wiper motor for sweeping moisture from the windshield, the windshield wiper control system comprising:
(a) a sensor disposed at a predetermined location on a vehicle windshield traversed by the wiper for generating an output indicative of the moisture impinging on the windshield proximate the predetermined location;
(b) smoothing means coupled to receive the sensor output, said smoothing means forming a weighted average of recent rain intensity;
(c) wiper mode control means operatively coupled to the smoothing means for establishing a sweep rate for the windshield wiper: and (d) means for electrically driving the wiper motor at the sweep rate determined by the wiper mode control means.

2. The rain responsive windshield wiper control system as in Claim 1 wherein the weighted average comprises means for combining recent rain intensity as measured by the sensor over a first predetermined time interval in a range of from about one to five seconds with that over a second predetermined time interval in the range of from about two to ten seconds.

3. A rain responsive windshield wiper control system for a vehicle of the type having a windshield, a windshield wiper adapted to be driven by a windshield wiper motor for sweeping moisture from the windshield, the windshield wiper control system comprising:
(a) sensor means disposed at a predetermined location on the vehicle windshield traversed by the wiper for generating rain event signals indicative of moisture impinging on the windshield proximate the predetermined location;
(b) means responsive to the rain event signals for repeatedly determining the frequency of occurrence of the rain event signals within a predetermined time interval to provide a recent rain flux value;
(c) smoothing means coupled to receive the recent rain flux value for eliminating momentary variations in the recent rain flux value;
(d) wiper mode control means operatively coupled to the smoothing means for establishing a sweep speed for the windshield wiper; and (e) means for electrically driving the wiper motor at the sweep speed determined by the wiper mode control means.

4. The rain responsive windshield wiper control system as in Claim 3 within the means for determining the frequency of occurrence of the rain event signals comprises means for totalizing a number of the rain event signals occurring within the predetermined time interval.

5. The rain responsive windshield wiper control system as in Claim 3 wherein the rain event signals comprise reversals of an output of the sensor means toward a predetermined quiescent level.

6. The rain responsive windshield wiper control system as in Claim 4 wherein the means for repeatedly totalizing a number of the rain event signals comprises:
(a) an analog-to-digital converter having an input connected to receive the rain event signals from the sensor means and an output;
(b) counting means connected to the output of the analog-to-digital converter for continuously counting the number of the rain event signals occurring within a given time period which is less than the predetermined time interval;
(c) means for summing the number of rain event signals occurring within successive ones of the given time periods over the predetermined time interval.

7. The rain responsive windshield wiper control system as in Claim 3 wherein the smoothing means comprises:
(a) means for forming a weighted average of the recent rain flux values over an extended time period greater than the predetermined time interval.

8. The rain responsive windshield wiper control system as in Claim 7 wherein the means for forming a weighted average of the recent rain flux values comprises:
(a) means for deriving a medium range average of the recent rain flux values;
(b) means for deriving a long range average of the recent rain flux values; and (c) means for combining the medium range average and long range average of the recent rain flux values.

9. The rain responsive windshield wiper control system as in Claim 3 wherein the wiper mode control means includes means for establishing a first rain intensity threshold at which the mode changes between an intermittent operation and slow continuous wiping speed, a second rain intensity threshold at which the mode changes between fast and slow continuous wiping speed and a third rain intensity threshold at which the mode changes between slow and fast continuous wiping speed, the third threshold being at a rain intensity level greater than that of the second threshold.

10. A rain responsive windshield wiper control system for a vehicle of the type having a windshield, a windshield wiper adapted to be driven by a windshield wiper motor for sweeping moisture from the windshield, the windshield wiper control system comprising:
(a) sensor means disposed at a predetermined location on the vehicle windshield traversed by the wiper for generating rain event signals indicative of moisture impinging on the windshield proximate the predetermined location:
(b) means responsive to the rain event signals for repeatedly determining the number of the rain event signals occurring within a predetermined time interval to provide a recent rain flux value;
(c) logarithmic conversion means for converting the recent rain flux value to a recent rain intensity value;
(d) smoothing means coupled to receive the recent rain intensity value for producing an output signal in which momentary variations in the recent rain intensity value are eliminated;
(e) wiper mode control means operatively coupled to the smoothing means for establishing a sweep rate of the windshield wiper; and (f) means for electrically driving the wiper motor at the sweep rate determined by the wiper mode control means.

11. The rain responsive windshield wiper control system as in Claim 10 wherein the rain event signals comprise reversals of an output of the sensor means toward a predetermined quiescent level.

12. The rain responsive windshield wiper control system as in Claim 10 wherein the means for determining the number of rain event signals occurring within a predetermined time interval comprises means for repeatedly totalizing a number of rain event signals occurring within discrete, equal time intervals whose sum equals the predetermined time interval.

13. The rain responsive windshield wiper control system as in Claim 12 wherein the means for repeatedly totalizing a number of the rain event signals comprises:
(a) an analog-to-digital converter having an input connected to receive the rain event signals from the sensor means and an output;
(b) counting means connected to the output of the analog-to-digital converter for continuously counting the number of the rain event signals occurring within a given time period which is less than the predetermined time interval: and (c) means for summing the number of rain event signals occurring within successive ones of the given time periods over the predetermined time interval.

14. The rain responsive windshield wiper control system as in claim 10 wherein the smoothing means comprises:
(a) means for forming a weighted average of the recent intensity values over an extended time period greater than the predetermined time interval.

15. The rain responsive windshield wiper control system as in Claim 14 wherein the means for forming a weighted average of the recent rain intensity value comprises:
(a) means for deriving a medium range average of the recent rain intensity value;
(b) means for deriving a long range average of the recent rain intensity value; and (c) means for combining the medium range average and the long range average.

16. The rain responsive windshield wiper control system as in Claim 10 wherein the wiper mode control means includes:
(a) means for establishing a first rain intensity threshold at which the wiping mode changes between an intermittent operation and a slow continuous wiping speed;
(b) a second rain intensity threshold at which the wiping mode changes between fast and slow continuous wiping speed; and (c) a third rain intensity threshold at which the wiping mode changes between a slow and a fast continuous wiping speed, the third threshold being at a rain intensity level greater than that of the second threshold.

17. The rain responsive windshield wiper control system as in Claim 10 wherein the wiper mode control means includes:
(a) exponential conversion means for converting the output signal from the smoothing means to a corresponding linear time value.

18. The rain responsive windshield wiper control system as in Claim 10 wherein the smoothing means comprises first and second filter means, each exhibiting an asymmetrical attack and decay time for increasing and decreasing recent rain intensity values, respectively.

19. The rain responsive windshield wiper control system as in Claim 10 wherein a time constant of the first filter means differs from a time constant of the second filter means.

20. The rain responsive windshield wiper control system as in Claim 15 wherein the means for deriving a long range average of the recent rain intensity includes:
(a) means for increasing a current value of the long range average of the recent rain intensity upon detection of an increase in setting of a driver-accessible sensitivity control; and (b) means for decreasing the current value of the long range average of the recent rain intensity upon detection of the decrease in setting of the drive-accessible sensitivity control.