Research On Key Technology Of The System Of The Motion Planning And Control With Applications To Autonomous Urban Driving

With the development of artificial intelligence,the autonomous driving attracts more and more attentions from the military,research institutes and other areas.For the autonomous vehicles,motion planning is one of its core technologies.The autonomous vehicle is faster and the applying environment is more complex and dynamic,which is very difficult for most existing robotic algorithms.This is especially true in the urban environment,complex conditions need to be handled separately.As a result,planning algorithms need to be more flexible in response to decision-making needs and the control algorithm should guarantee the small tracking error and smooth control process.This paper proposes a planning algorithm which accommodates the autonomous driving based on the asymptotically optimal sampling-based algorithm,anytime CL_SST.For the vehicle control,this paper presents a new lateral controller,CF-Pursuit and a longitudinal controller based on the fuzzy logic.Specific research content is listed as follows:1)This paper discusses the domestic and international research developments on the autonomous vehicles,compares the path planning and lateral control of recent years.Based on the specific requirements of the urban driving environment,we propose the key planning and control problems in urban driving.2)The "Tuzhi Ilutonomous vehicle is introduced,including the components of the system,hardware and software compositions.Moreover,we design a decision system based on the finite state machine.3)The autonomous driving planning method is studied based on the asymptotically optimal sampling-based algorithm.Combining with the demand of urban driving environment,we design a planning algorithm on the basis of the asymptotically optimal sampling-based algorithm SST,anytime CL_SST.The "anytime" attribute could better guarantee the real-time computation.CL_SST combines closed-loop strategy with 6 D vehicle motion model and the asymptotically optimal attribute of SST to have the good anti-jamming property and the optimal characteristics.Based on the SST,CL_SST improves the sampling,parent-node selection and extention strategies.Moreover,we build an objective function to select the path to accommodate with various driving conditions.4)The control system is built by designing the lateral and longitudinal controllers.A lateral controller CF_Pursuit is designed based on the geometric controllers.We analyze the principle of the geometric controllers and make use of the Clothoid fitting method to replace the circle to improve the fitting error and decrease the tracking error.Then,we design a look-ahead distance tuning strategy by combining the maximum curvature of the fitting curve and the fuzzy logic.Moreover,we design a longitudinal controller based on the fuzzy logic.The comparing experimental results in the open road show that the lateral controller CF Pursuit can meet the practical requirements of urban driving in path tracking,and can maintain the cross-track error at 0.5m under 30 km·h-1.5)The whole system is tested in real urban roads,including long-distance test,static obstacle avoidance,dynamic obstacles follow and overtake and the complex scenario with static and dynamic obstacles.Experimental results demonstrate that the proposed planning and control system could assist the "Tuzhi ?" autonomous vehicle perform well when driving at a maximum of 36 km·h1-1.During the obstacle-avoidance test,the autonomous vehicle can run at 30 km·h-1 when keeping lane and 15 km·h-1 when avoiding the obstacle.When following the front car,the autonomous vehicle can accommodate the velocity required to maintain a reasonable distance and smooth driving.During the complex test scenario,the autonomous vehicle passes the test with safe and smooth driving by the interaction between the decision and the planning systems.