基于LSTM的自动驾驶车道预测与决策方法研究

Study on Autonomous Lane Prediction and Decision Method Based on LSTM

自动驾驶汽车是促进智能交通系统发展的重要工具,能够有效减少人为因素引发的交通事故,减少道路拥挤造成的能源消耗,具有重要的理论意义和实用价值。决策控制作为自动驾驶汽车的关键共性技术,主要负责对环境信息理解和推理并得出合理驾驶行为,生成可行域内最优轨迹,最后控制车辆执行机构执行相应动作。本文针对复杂多变的动态场景下交通车辆轨迹预测与自动驾驶汽车换道的预测与决策方法问题进行分析,提出一种基于长短期记忆网络(Long Short-Term Memory,LSTM)的自动驾驶车道预测与决策方法,从而有效提高自动驾驶汽车车道预测与决策的合理性、安全性和智能性,促进在复杂动态场景下自动驾驶汽车智能化水平的提升。

Autopilot is an important tool to promote the development of intelligent transportation system. It can effectively reduce traffic accidents caused by human factors and reduce energy consumption caused by road congestion. It has important theoretical and practical value. Decision control, as the key common technology of automatic driving vehicle, is mainly responsible for understanding and reasoning about environmental information, obtaining reasonable driving behavior, generating optimal trajectories in feasible area, and finally controlling vehicle actuators to perform corresponding actions. In this paper, we analyze and study the prediction and decision-making methods of vehicle traffic trajectory and driving lane change in complex and dynamic scenarios, and propose an automatic driving lane prediction and decision-making method based on LSTM, which can effectively improve the rationality of lane prediction and decision making for automatic driving vehicles. Safety and intelligence promote the intelligent level of automatic driving in complex dynamic scenes.