航空学报 > 2017, Vol. 38 Issue (11): 321275-321275   doi: 10.7527/S1000-6893.2017.321275

基于GA-OCPA学习系统的无人机路径规划方法

刘鑫1, 杨霄鹏1, 刘雨帆2, 姚昆1   

  1. 1. 空军工程大学 信息与导航学院, 西安 710077;
    2. 北京航空航天大学 电子信息工程学院, 北京 100083
  • 收稿日期:2017-03-27 修回日期:2017-07-23 出版日期:2017-11-15 发布日期:2017-07-21
  • 通讯作者: 刘鑫 E-mail:kdyliuxin@163.com
  • 基金资助:

    国家自然科学基金(61202490);航空科学基金(20150896010)

UAV path planning based on GA-OCPA learning system

LIU Xin1, YANG Xiaopeng1, LIU Yufan2, YAO Kun1   

  1. 1. Information and Navigation Institute, Air Force Engineering University, Xi'an 710077, China;
    2. School of Electronics and Information Engineering, Beihang University, Beijing 100083, China
  • Received:2017-03-27 Revised:2017-07-23 Online:2017-11-15 Published:2017-07-21
  • Supported by:

    National Natural Science Foundation of China (61202490); Aeronautical Science Foundation of China (20150896010)

摘要:

为解决未知空域中无人机路径规划方法实时性和适用性不足的问题,以生物应激条件反射理论为基础,将无人机实时路径规划类比为在外界条件刺激下的一种自学习行为。首先,将概率自动机与遗传算法相结合,设计了基于Skinner操作条件反射理论框架(GA-OCPA)的学习系统;然后,将无人机规避机动的飞行速度、滚转加速度和拉升加速度作为系统学习的行为,并计算每次学习尝试之后的选择概率和个体适应度,通过遗传算法搜索最优行为进而得到最优路径;最后,运用增量多层判别回归树(IHDR)对学习得到的最优行为建立知识库,形成威胁状态与路径规划的匹配映射。实验结果表明GA-OCPA学习系统对于无人机路径规划具备有效性和适用性。

关键词: 无人机, 路径规划, 遗传算法, 操作条件反射, 概率自动机

Abstract:

To solve the problem of deficiency in real-timeliness and applicability of path planning for the Unmanned Aerial Vehicle (UAV) in the unknown airspace, the real-time path planning of the UAV is simulated as a self-learning behavior under the condition of external stimuli, based on the biological operant conditioning theory. The probabilistic automaton is combined with the genetic algorithm to construct a learning system of Genetic Algorithm-Operant Conditioning Probabilistic Automaton (GA-OCPA) according to the Skinner operant conditioning. The UAVs' evasion maneuvering flight speed, rolling acceleration and climbing acceleration are taken as the learning behaviors of the system, and the probability of selection and individual fitness are calculated after each learning attempt. The optimal path can then be obtained by searching for the best behavior using the genetic algorithm. The knowledge base of the best learned behaviors is established using Incremental Hierarchical Discriminant Regression (IHDR), and the matching mapping between the threat state and path planning is then formed. The result shows the viability and applicability of the GA-OCPA learning system for UAV path planning.

Key words: Unmanned Aerial Vehicle (UAV), path planning, genetic algorithm, operant conditioning, probabilistic automaton

中图分类号: