基于强化学习的循迹小车实现

Realization of Track-and-trace Car Based on the Reinforcement Learning

近年来,强化学习在棋类游戏、视频游戏、复杂机械控制、智能决策等领域大放异彩。强化学习以其较少的人工劳动量,易于调参和较为稳定的特性而逐步受到广大科研人员和企业的青睐。强化学习也是笔者所在实验室的研究方向之一,本文在指导教师及强化学习相关理论的指导下,结合之前对树莓派等智能硬件的一些了解,尝试了使用强化学习方法来实现现实中小车循迹的功能,收到了较为不错的效果。本论文共分四部分:系统总体设计,软、硬件的设计和实现,实验结果和分析以及结论与展望。

In recent years,reinforcement learning has flourished in the fields of board games,video games,complex mechanical control,and intelligent decision-making.Reinforcement learning is gradually favored by researchers and enterprises because of its less labor,easier adjustment and more stable characteristics.Reinforcement learning is also one of our research directions.Under the guidance of instructors and theories of reinforcement learning,this paper combines some previous knowledge of smart hardware such as Raspberry Pi.We try to use reinforcement learning method to realize the function of car tracking in reality,and get a good effect.This paper is divided into four parts:system overall design,hardware and software design and implementation,experimental results,conclusions and prospects.