Development of a Trajectory Planner for Automated Low Speed Functions

Techincal University of Darmstadt (TUDa) is collaborating with several German universities and industry partners in a project called AUTOtech.agil, which aims to create an open architecture for the mobility system of the future. One of the key elements of the AUTOtech.agil project is the low-speed function, which is an automatic driving function for low-speed driving conditions. Among the low-speed functions there is an important safety feature called ”Safe Halt”, which uses redundant sensors to perceive environmental information and bring the vehicle safely to a stop in case of emergency. It is possible to extend the range of application of the ”Safe Halt” function, and one important task is to develop a trajectory planner based on the ”Safe Halt” function. The trajectory planner is used in the low-speed function and is designed to achieve low-speed, small-area usage scenarios for path finding. The purpose of this thesis is to develop the trajectory planner for the low-speed function. This thesis presents the design and implementation process of a trajectory planner programmed in Python. The trajectory planner takes as input the start and end positions, as well as the static obstacle map, and will output a trajectory with a speed profile. To achieve the functionality described above, a suitable path finding algorithm as well as the ability to adjust the vehicle’s pose based on the non-holonomic constraints of the vehicle are required. The trajectory planner uses the Hybrid A* algorithm as the path finding algorithm, which is an improved version of the Regular A* algorithm. The Hybrid A* algorithm takes into account the non-holonomic constraints of the vehicle and improves the generation process of trajectories through cost functions. The Reed-Shepp curve is used to adjust the pose of the vehicle so that it reaches its destination in the specified orientation. After the initial trajectory has been generated, the trajectory will also be post-processed. The post-processing function includes a gradient descent smoother and an interpolator, which further improve the quality of the trajectory. Finally, the generated trajectory will also be subjected to speed profile planning, which assigns a reasonable and safe speed profile to the trajectory. After the implementation progress, the functionality of the trajectory planner will be verified in both Python and simulation environments. The trajectory planner presented in this thesis is well suited to the design requirements for trajectory planning in static environments. It offers good functionality, safety, and overall performance.