Threat Assessment And Trajectory Planning For Intelligent Vehicle Obstacle Avoidance System

The number of traffic accidents remains high with the high speed of vehicle ownership expansion; vehicle safety has attracted increasing attention and promoted the research and development on vehicle active safety. In the vehicle active safety research, obstacle avoidance system is one of the important researches with great significance in reduce collision.The basic approach of the vehicle obstacle avoidance system is based on the known surroundings to assess the vehicle threat when driving along the current states, if the result is dangerous, then assesses the possible obstacle avoidance maneuvers(steering or braking) and selects the maneuver with the smallest threat as the obstacle avoidance maneuver, according which planning a safe obstacle avoidance trajectory, and finally the trajectory is tracked by the controller.Currently, threat assessment for obstacle avoidance system mostly deterministic predicts future trajectories of the obstacles and host vehicle, less considering the uncertainty of future trajectory, the result can be low reliability. A small number research considering predicted trajectory uncertainty is difficult to meet real-time requirements because of the computational complexity, and less considering the case of braking while steering maneuver to avoid the obstacles. Despite there are many trajectory planning methods, how to adapt to complex environment dynamic changes, explore the feasible region as much as possible to find better trajectory and many other aspects need to be further studied.This thesis focuses on the threat assessment and trajectory planning for intelligent vehicle obstacle avoidance system. In the threat assessment phase, combining the deterministic and probabilistic trajectory prediction to improve the calculation speed simultaneously ensure the authenticity; considering the case of braking while steering maneuver to avoid the obstacles to expand the searching region which is more consistent with the real traffic environment. In the trajectory planning phase, uses the rapidly exploring random tree(RRT) to plan trajectory, in views of the problems existing in the RRT algorithm and based on the existing improved RRT algorithms, the thesis improved the goal biasing and node connection for the random tree expansion and proposed the post processing mechanism for the random tree including node pruning and trajectory smoothing to generate more security trajectory to avoid obstacle.The main research work in this thesis includes:1) Threat assessment for obstacle avoidance system. The thesis assumes that already knows the driving intention of the transport vehicles, focus on the transport vehicles trajectory prediction, host vehicle obstacle avoidance trajectory cluster assessment and collision probability calculation. Combines the deterministic and probabilistic trajectory prediction to improve the calculation speed simultaneously ensure the authenticity, if the distance between the obstacle and host vehicle trajectories is too close, considering the probabilistic of the predicted trajectories. Considering the case of braking while steering maneuver to avoid the obstacles to expand the feasible region when assess the host vehicle obstacle avoidance trajectory cluster. To improve the calculation speed using the improved Monte Carlo simulation method to calculate the collision probability.2) Preliminary research on decision making in structural road for avoid obstacle. Decision making is selecting the best maneuver to avoid obstacle, in the thesis, studied two typical scenarios, lane and intersection driving. Given the limited research time, the decision making is relatively simple.3) Trajectory planning for obstacle avoidance system. The thesis uses the RRT algorithm to plan the trajectory. To ensure effectiveness and computational speed, there are many improvements to generate smooth trajectory as well as satisfy vehicle dynamic constraints. Due the randomness of the results, the algorithm generates many feasible trajectories and selects the best one according to obstacle avoidance requirement and stability. The trajectory track controller is designed using optimal preview control.4) Simulation test for the obstacle avoidance system. The simulation results show that the proposed algorithm and system work as expected and is valid and effective, able to handle complex environment dynamic changes.