基于MDP的无人机避撞航迹规划研究

Research on UAV Collision Avoidance Path Planning based on MDP

摘要:

无人机基于避撞条件下的目标搜索航迹规划是指针对复杂且众多的环境障碍约束通过合理规划飞行路径,以更快、更高效地形式找到目标；首先深入探讨了概率论视角下有限位置马尔科夫移动的规律,构建了相应的马尔科夫移动分布模型；随后在借鉴搜索系统航迹规划的前沿研究成果之上,结合马尔科夫决策过程理论,创新性地引入了负奖励机制对Q-Learning策略算法迭代,构建了单无人机目标搜索模型；并通过类似“风险井”的可视化方式直观地将障碍约束对飞行影响呈现出来；最后进行仿真实验论证算法的可行性和有效性,对航迹规划算法的设计具有一定的参考意义。

Abstract:

Target search path planning based on collision avoidance of UAV is to find the target in the faster and more efficient form by reasonable flight path planning against complex and numerous environmental obstacles. Firstly, this paper deeply discussed the law of finite position Markov mobility from the perspective of probability theory, and constructed the corresponding Markov mobility distribution model. Then, based on the cutting-edge research results of search system trajectory planning, combined with the Markov decision process theory, the negative reward mechanism was innovatively introduced to iterate the Q-Learning strategy algorithm, and the single UAV target search model was constructed. And the impact of obstacle constraints on flight is visually presented through a visualization method similar to "risk well". Finally, the simulation experiment proves the feasibility and effectiveness of the algorithm, which has certain reference significance for the design of the route planning algorithm.