摘要
为提高无人机在观光农业混合型无线传感器网络中的数据采集效率,提出一种基于深度强化学习的无人机路径规划策略。基于社会力模型引入人流参量,结合Semi-Markov-Option分层强化学习方法以降低模型复杂度,基于Rainbow算法提出SMO-Rainbow(Semi-Markov-Option-Rainbow)路径规划策略。在ε-greedy探索策略中引入Tanh函数,提出AT-ε-greedy(adaptive-Tanh-greedy)策略,进一步平衡深度强化学习模型训练中的探索与利用阶段。实验结果表明,在观光农业场景中,所提路径规划策略与其它深度强化学习无人机路径规划策略相比,数据采集效率与训练稳定性均更优,有效降低了模型训练难度。
To improve the efficiency of UAV data collection in wireless sensor networks for tourism agriculture,a UAV path planning strategy based on deep reinforcement learning was proposed.The human flow parameters were introduced based on the social force model,the SMO-Rainbow(Semi-Markov-Option-Rainbow)path planning strategy was proposed based on the Rainbow algorithm by combining the Semi-Markov-Option hierarchical reinforcement learning method to reduce the model complexity.The Tanh function was introduced in theε-greedy exploration strategy,and the AT-ε-greedy(adaptive-Tanh-greedy)strategy was proposed to further balance the exploration and exploitation phases in the training of deep reinforcement learning models.Experimental results show that the proposed path planning strategy has better performances in terms of data collection efficiency as well as training stability compared with other deep reinforcement learning UAV path planning strategies in the scenario of tourism agriculture,and effectively reduces the model training difficulty.
作者
张凡
万雪芬
崔剑
刘会丹
蔡婷婷
杨义
ZHANG Fan;WAN Xue-fen;CUI Jian;LIU Hui-dan;CAI Ting-ting;YANG Yi(College of Information Science and Technology,Donghua University,Shanghai 201620,China;Hebei IoT Monitoring Engineering Technology Research Center/College of Computer,North China Institute of Science and Technology,Langfang 065201,China;School of Cyber Science and Technology,Beijing University of Aeronautics and Astronautics,Beijing 100083,China)
出处
《计算机工程与设计》
北大核心
2022年第7期1905-1914,共10页
Computer Engineering and Design
基金
国家重点研发计划基金项目(2018YFC0808306)
廊坊市科学技术研究与发展计划基金项目(2019011010)
秦皇岛市科学技术研究与发展计划基金项目(201805A016)
河北省物联网监控工程技术研究中心基金项目(3142018055、3142016020)。
作者简介
张凡(1996-),男,江苏连云港人,硕士研究生,CCF学生会员,研究方向为物联网技术;万雪芬(1979-),女,江苏连云港人,硕士,副教授,CCF会员,研究方向为物联网技术;崔剑(1980-),男,北京人,博士,讲师,研究方向为嵌入式系统;刘会丹(1997-),女,河南开封人,硕士研究生,研究方向为物联网技术;蔡婷婷(1998-),女,河南漯河人,硕士研究生,CCF学生会员,研究方向为深度学习;通讯作者:杨义(1978-),男,四川成都人,博士,副教授,硕士生导师,CCF会员,研究方向为物联网技术及智慧农业。E-mail:yiyang@dhu.edu.cn。
