基于DDPG算法的矿用电机车防撞及避障策略研究  

Research on Collision Avoidance and Obstacle Avoidance Strategy of Mining Electric Locomotive Based on DDPG Algorithm

作　　者：鲁良伟 李敬兆[1,2] 周华平 LU Liangwei;LI Jingzhao;ZHOU Huaping(School of Electrical and Information Engineering,Anhui University of Science and Technology,Huainan 232001,China;School of Computer Science and Engineering,Anhui University of Science and Technology,Huainan 232001,China)

机构地区：[1]安徽理工大学电气与信息工程学院,安徽淮南232001 [2]安徽理工大学计算机科学与工程学院,安徽淮南232001

出　　处：《煤炭技术》2022年第2期184-188,共5页Coal Technology

基　　金：安徽省重点研发计划、国际科技合作专项(202004b11020029);国家自然科学基金项目(51874010,61170060);北京理工大学高精尖机器人开放性研究项目(2018IRS16);物联网关键技术研究创新团队(201950ZX003)。

摘　　要：为保障矿用电机车的安全稳定运行,实现电机车运输过程中避障的自主化与智能化,提出了一种基于深度确定性策略梯度(DDPG)算法的电机车防撞及避障策略。通过融合电机车行驶环境中的多源特征数据,作为DDPG算法网络的状态输入,结合奖励函数、OU噪声来训练智能体,进而输出确定性防撞避障策略至电机车整车控制器,以控制电机车完成转向、调速、制动及报警等避障动作。仿真测试结果表明,所设计的避障防撞策略有效解决了电机车在井下复杂工况中的运行控制问题,对电机车的少人、无人化驾驶研究具有一定的意义。In order to ensure the safe and stable operation of mining electric locomotive and realize the autonomy and intelligence of obstacle avoidance in the transportation process of electric locomotive,an anti-collision and obstacle avoidance strategy of electric locomotive based on depth deterministic strategy gradient(DDPG)algorithm is proposed.By fusing the multi-source characteristic data in the motor vehicle driving environment as the state input of the DDPG algorithm network,combined with the reward function and OU noise to train the agent,and then output the deterministic anti-collision and obstacle avoidance strategy to the motor vehicle controller to control the motor vehicle to complete obstacle avoidance actions such as steering,speed regulation,braking and alarm.The simulation test results show that the designed obstacle avoidance and collision avoidance strategy effectively solves the operation control problem of electric locomotive under complex underground working conditions,and is of certain significance to the research on few people and unmanned driving of electric locomotive.

关 键 词：矿用电机车 DDPG算法 防撞避障策略 OU噪声 智能体 

分 类 号：TD67[矿业工程—矿山机电]