Decision making and control
In the aspect of intelligent decision theory, the path and motion integration optimization method is proposed. Based on the classical Pontryagin's Greatest Value Theory, an improved method is proposed to relax the gear and lane control variables into a batch of continuous variables, and the collaborative optimization of the software system, hardware system, and road integration environment is also considered to realize integrated optimization, and a model framework is built to accelerate the solution by designing artificial gradients. The iterative intra-local quadratic fitting solution method is proposed and its computing efficiency is much higher than that of Newton's method. This algorithm reduces the response time from seconds to 0.05 seconds with guaranteed convergence, robustness, and efficiency. This research has been highly regarded by the U.S. Department of Transportation and the U.S. Department of Energy, and has been recognized by the U.S. Department of Transportation as a "Nationwide Significant Contribution".
The method has been applied in some practical projects.
1.AVP decision making and control under Vehicle-to-Infrastructure cooperation system（Cooperator:Renault S.A.,QCraft）
Considering the real-time static / dynamic obstacles and taking the parking slot as the spatial constraint, plan the trajectory from the alighting point (such as company, community, etc.) to the parking slot; On this basis, based on the optimization control methods, taking the static / dynamic obstacle information around the vehicle as the constraint, the accurate tracking of the track is realized; Finally, the AVP system is evaluated from the perspectives of safety, accuracy and arrival rate, so as to optimize the algorithm design and simulation test at the same time.
2.Automated lane change controller for L2 autonomous vehicle (Cooperator: Tenew Automotive Technology (Suzhou) Co., Ltd)
This program develops an automated lane-change controller for autonomous driving scenarios when speed is greater than 120 km/h. It is with the following features: 1) with route planning and motion planning integrated; 2) considering vehicle tire and steering dynamics; 3) with capability of lane-change in high-speed traffic scenarios.
3.Path tracker for automated vehicles (Cooperator: ECARX Co., Ltd)
This program develops a path tracker for automated vehicles. This path tracker is based on a model predictive control method in a mixed domain. It is longitudinal and lateral planning decoupled. Field tests have been conducted in straight-cruising, turn, and u-turn scenarios. Results show that the control error is less than 12 cm, which confirms the proposed method reaches world leading level.