基于深度强化学习的多智能体高速车道检测研究

Multi-agent High-speed Lane Detection Based on Deep Reinforcement Learning

随着汽车技术与人工智能领域的发展,高级驾驶员辅助系统越来越流行,自动驾驶技术能够极大地提高效率、降低成本。相对应的可行驶区域检测、车道线检测和车速估计等便成为了关键性的一环。一些检测方法仅检测带有粗略边界框的车道线,而忽略特定曲线车道的形状,将深度强化学习引入到了车道检测模型中,采用智能体定位点移动来实现准确的车道检测和定位。用以改善辅助驾驶系统,提高车道线识别情况。无论是对于驾驶员还是自动驾驶系统,都有很好的驾驶体验改善作用。

The detection of relative driving areas,lane line detection,and speed estimation have become crucial components.Some detection methods only detect lane lines with rough bounding boxes,ignoring the shape of specific curved lanes.This paper introduces deep reinforcement learning into the lane detection model,using intelligent agent positioning point movement to achieve accurate lane detection and positioning.To improve the auxiliary driving system and improve the recognition of lane markings.Both the driver and the auto drive system can improve the driving experience.