Research On Human-simulated Decision And Control Algorithm For Parking Assist System

Parking assist system (PAS) is an important component of advanced driver assistancesystems (ADAS), and an integral component of unmanned driving technology. Drivers couldpark safely, accurately, and quickly with the help of PAS, which designed to reduce the rateof traffic accident due to misoperation and avoid unnecessary loss. Meanwhile, thedevelopment and promotion of PAS will not only play a significant role in unmanned vehicletechnology applications, but also adapt to the development of vehicle electronic controlsystems in the future.The major role of PAS is to replace the driver in direction handling, assisting to park acar into the designated lot during parking process by the environment awareness and vehiclestatus. As a new rising system of ADAS, PAS is a requisite for developments of intelligentvehicle. Nowadays, most R&D institutions developed it as an independent electronic unit,which leads to the plight of integrating with other advanced chassis electronic control systemone day. At the same time, the use of high-cost sensors increases great resistance toproduction application.This dissertation has conducted a research on human-simulated decision&controlalgorithm for PAS, in order to integrate with existing driver model theory, which aims toextend driver model to backwards parking mode. In the meantime, under the condition ofremaining current sensor devices equipped on the tested vehicle unchanged, it aims to applythe low-cost PAS, which realizes assisting the tested car parking.After collecting and analyzing drivers’ parking operations, this article provides aparking trajectory dynamic decision method based on fuzzy analytic hierarchy process(FAHP) and PID controller of backward vehicle direction, forms the PAS that based on thehuman-simulated decision&control algorithm. According to the multi-condition test ofCarSim/LabVIEW co-simulation, the designed PAS matches the tested vehicle, which leadsto a successful parking.The major contents of this research can be summarized as follows:Firstly, the study is built on the analysis of the driver parking operation behaviorcharacteristics, which is regarded as the theoretical foundation for designing the unit of human-simulated decision&control. In order to seek the common characteristics of thedriver parking operation, it designs and completes the experience on recording the operationdata about driver parking behavior. Via CAN bus, the acquisition program stores dataperceived by vehicle sensors. Compared with the subjective evaluation of parking indirectparticipators and referring to parking experience, general rules of parking operation could besummed up, based on which to draw the reference parking path.Secondly, referring to the results of analyzing drivers’ parking behavior characteristics,it puts forward the framework of the decision&control simulated human. According to thetheory of preview-follower dynamic correction driver model and the regular rules of parking,it turned to accept the FAHP method to judge and choose the future track points. Specificallyspeaking, the introduction of safety, consistency and efficiency fuzzy evaluation indexesmake it possible to make dynamic trajectory decision and figure out corresponding optimalpreview lateral acceleration. Based on low-speed direction driver model, it is easy to build aparking direction control mode via PID controller, and optimize the PID parameters usinggenetic algorithm by introducing quadratic performance indexes of vehicle location andheading angle deviation.Finally, this article has conducted multi-condition parking experiments in CarSimsimulation environment to verify the feasibility of the decision&control simulated humansystem. In the application of actual vehicle parking, Kalman filtering is adopted to reduce thenoise of wheel speed signals and then vehicle speed could be estimated by wheel speed.Adding the estimation of vehicle yaw rate, the vehicle state estimation model isaccomplished. It’s necessary to rewrite PAS program for tested vehicle and display interfaceusing LabVIEW. Through conducting parking experiences on tested vehicle equipped withthis program, it proves that the designed PAS works and leads to a successful parkingultimately.