IEEE IEEE INTELLIGENT SYSTEMS Published by the IEEE Computer Society Intelligent Transportation - PDF document

internal vehicle computer. We designed ourdriving area to emulate an urban environmentbecause automatic urban driving is one ofITSÕs less researched topics.We modeled the guidance system usingfuzzy variables and rules. In addition to thesteering wheel and vehicle velocity func-tionalities,we also consider variables that thesystem can use in adaptive cruise control(ACC) and overtaking capabilities. Amongthese variables are the distance to the nextbend and the distance to the lead vehicle (thatis,any vehicle driving directly in front of theautomated vehicle).Car driving is a special control problemcomplex and canÕt be accurately linearized.We use fuzzy logic because itÕs a well-testedprovides good results,and can incorporatehuman procedural knowledge into controlalgorithms. Also,fuzzy logic lets us mimichuman driving behavior to some extent.Steering controlThe steering control systemÕs objective isto track a trajectory. To model lateral andangular tracking deviations perceived by ahuman driver,we use two fuzzy variables:. These variablesrepresent the difference between the vehi-cleÕs current and correct position and its ori-entation to a reference trajectory.Both variables can take values. between the orientation and vehicle velocityvectors. If this angle is counterclockwise,thevalue is wise,the value is represents the distance from the vehicle to thereference trajectory. If the vehicle is positionedon the trajectoryÕs left,the value is; itÕs if the vehicle is on the right.We compute the variablesÕinstantaneousvalue using the DGPS data and a digital envi-ronment map. The fuzzy output variable iscorrect the input errors. Again,the variablelinguistic values. Thevalue is counterclockwise,and clockwise. We define the fuzzy sets thatdefine the values in an interval ofÐ540 degrees and 540 degrees.As with human behavior,our guidancesystem works differently for tracking lanesor turning on sharp bends. When travelingalong a straight road,people drive at rela-tively high speeds while gently turning thesteering wheel. In contrast,on sharp bends,they rapidly reduce speed and quickly turnthe steering wheel. We emulate such behav-ior by changing the membership function,and linguistic variables. Torepresent the human procedural knowledgein the driving task,we need only two fuzzyrules. These rules tell the fuzzy inferencemotor how to relate the fuzzy input and out-put variables:Although these rules are simple,they gen-erate results that are close to human driving.The rules are the same for all situations,butthe definition of the fuzzy variablesÕlinguis-tic values change. Figure 3 shows this fea-ture in the membership function definition3b show the degree of truth for the input errorvalues in straight-path tracking situations.This definition lets the system act quicklywhen trajectory deviations occurÑagain inkeeping with human behavior.To prevent accidents,we must limit thedriving. This limitation is also similar tohuman behavior; we achieve it by definingthe output variable membership function asa singleton,confining this turning to 2.5 per-cent of the total. Figure 3c and 3d show sim-ilar function definitions,but their shapeÕs gra-dient is lower. This makes the driving systemless reactive when tracking a straight trajec-tory and assures that theyÕll adapt to the routesmoothly. We can also represent the output www.computer.org/intelligentIEEE INTELLIGENTSYSTEMS Figure 1. The Atestbed vehicles. An embedded fuzzy-logic-based control systemcontrols both speed and steering in each Citro‘n Berlingo. Guidance system Steering wheelThrottleBrake Tachometer LAN Camera motor Analog card motor Vision system Figure 2. The Asystem control structure. The sensorial equipment supplies thecontrol signals to manage the vehicle actuators (steering wheel, throttle, and brake ). We fine-tuned the membership functionsexperimentally,comparing their behaviorceptably. So,the driving system selects athe situation,which leads to different reac-tions for each route segment.Speed controlTo control speed,we use two fuzzy inputvariables:. To con-trol the accelerator and the brake,we usetwo fuzzy output variables:crisp value is the differencebetween the vehicleÕs real speed and theuser-defined target speed,and the crisp value is the speedÕs variation during atime interval. The throttle pressure range is2Ð4 volts,and the brake pedal range is0Ð240 degrees of the actuation motor.knowledge for throttle control areThe rules for brake control aremeans depress the brakeand throttle,and brake/throttle up the brake and throttle. The associated mem-define the degree of nearness to 0,respectively.Figures 3e through 3h show the member-(for the brake controller) for,respectively. Anasymmetry exists in the two variable defini-tions for two reasons:¥to account for the difference in how accel-erating and braking vehicles behave,andJANUARY/FEBRUARY 2007www.computer.org/intelligent (b)(a)(d)(c)(f)(e)(h)(g) –0.80 –20Right LeftRight Left –180Degrees –150 null –350 null –5.90 –530 –180180180Degrees –140LESS THAN nullfMORE THAN nullf nullf –50 nullf Figure 3. The membership function definition for fuzzy variables: (a) r throttle, (f) throttle, (g)