快速发展的汽车行业给人民群众带来出行便利的同时，也使得交通事故的危害程度日渐提升。随着人工智能软件、硬件技术的协同发展，利用自动驾驶技术减少由驾驶员人为因素引发的交通事故的方案逐渐走向现实。路径规划是自动驾驶技术的关键一环，结合实时交通信息和车辆物理信息，对预先规划的行车路线进行重规划。如何保证计算资源受限、计算时间受限场景下的车辆避障安全是路径规划的研究难点之一。针对上述车辆避障路径规划问题，本文提出了一种基于量子强化学习的车辆避障路径规划算法。自动驾驶车辆可以通过获取车载超声波传感器数据，实现对行驶道路信息和障碍物信息的提取，并基于量子强化学习和多层控制策略，动态输出车辆行驶轨迹，完成车辆避障任务。论文的主要研究成果包括：1. 提出了基于混合量子-经典计算框架的量子强化学习算法，利用变分量子电路的量子纠缠特性，解决了强化学习在资源受限场景中的应用问题，大幅度地减少了算法模型的参数量。2. 提出了面向车辆避障问题的路径规划算法，设计了基于路径规划算法的车辆轨迹控制器，改进人工势场的目标引力场模型，解决了车辆、目标点和障碍物接近时运动轨迹剧烈抖动的问题。3. 提出了基于量子强化学习的多层控制策略，整合量子强化学习和经典路径规划算法，解决了经典路径规划方法在加权的安全性指标下表现较差的问题，减少模型参数量的同时，提升了算法的性能。本文围绕车辆避障的路径规划问题，搭建了匀速行驶车辆在单向双车道上让行人的仿真环境并进行数值仿真实验。实验结果表明，量子强化学习的模型参数量降低为经典强化学习的 1/45，并提升 15.8% 的安全性平均得分。本文提出的算法解决了在资源受限的场景中自动驾驶车辆的行驶安全问题。

While the rapidly developing automotive industry has brought convenience to the people, it has also made traffic accidents increasingly dangerous. With the synergistic development of artificial intelligence software and hardware technologies, the use of autonomous driving technology is becoming a reality to reduce the number of accidents caused by human factors of drivers. Route planning is a key part of autonomous driving technology, combining real-time traffic information with physical information about the vehicle to re-plan pre-planned driving routes. How to ensure the safety of vehicle obstacle avoidance in scenarios with limited computing resources and limited computing time is one of the research difficulties in path planning.To address the above-mentioned vehicle obstacle avoidance path planning problem, this paper proposes a quantum reinforcement learning-based vehicle obstacle avoidance path planning algorithm. The self-driving vehicle can achieve the extraction of driving road information and obstacle information by acquiring the on-board ultrasonic sensor data. And based on quantum reinforcement learning and multi-layer control strategy, the vehicle driving trajectory is dynamically outputted to complete the vehicle obstacle avoidance task. The main research contents of this paper are shown in the following aspects. 1. A quantum reinforcement learning algorithm based on a hybrid quantum-classical computing framework is proposed, which exploits the quantum entanglement property of the variational quantum circuits. The problem of applying reinforcement learning in resource-constrained scenarios is solved, and the number of parameters of the algorithm model is significantly reduced.2. A path planning algorithm is proposed for the vehicle obstacle avoidance problem, and a vehicle trajectory controller based on the path planning algorithm is designed. The target gravitational field model of the artificial potential field is improved to solve the problem of violent jittering of the motion trajectory when the vehicle, target point and obstacle approach.3. This paper proposes a multi-layer control strategy based on quantum reinforcement learning, integrating quantum reinforcement learning and classical path planning algorithms. The problem of poor performance of the classical path planning method under the weighted safety index is solved. The performance of the algorithm is improved while reducing the number of model parameters.This paper focuses on the path planning problem of vehicle obstacle avoidance. In this paper, a simulation environment is built to avoid pedestrians in a one-way two-lane carriageway with uniform speed. Numerical simulation experiments are also conducted. The experimental results show that the number of model parameters for quantum reinforcement learning is reduced to 1/45 of classical reinforcement learning, and the average safety score is improved by 15.8%. The proposed algorithm solves the driving safety problem of self-driving vehicles in resource-constrained scenarios.