Research On Unmanned Aerial Vehicle Path Planning Based On Intelligent Algorithms

UAVs have been widely used in both military and civilian fields due to their exceptional maneuverability and adaptability to harsh environments.In particular,UAVs are often used for complex and varied tasks in the military field.However,more complex environments require higher demands on UAV.For instance,when a UAV flies at a fixed altitude,it needs to avoid threats such as missiles,artillery,and electromagnetic interference.Similarly,when a UAV performs tasks in more complex three-dimensional unknown environments,it needs to perceive threats in real-time and re-plan their flight paths when a threat is detected,which puts higher demands on the real-time performance of path planning.Additionally,when multiUAV perform tasks in unknown environments,they need to avoid threats and obstacles that may appear at any time,as well as avoid collisions with each other.The purpose of this thesis is to investigate path planning problems for UAV in various environments.To address these issues,we employed intelligent algorithms.Specifically,we tackled path planning problems for UAV in two-dimensional known environments,threedimensional unknown environments,and multi-UAV path planning problems in threedimensional unknown environments.Our research work takes into account the constraints of the drone’s flight environment and its own performance.Further details on our research findings can be found below.(1)For the problem of UAV path planning in two-dimensional known environments,this thesis proposes an improved artificial bee colony algorithm(Improved Artificial Bee Colony Algorithm Integrating Grey Wolf Optimization Algorithm,GWOABC)that integrates the grey wolf algorithm.By incorporating the search rules of the grey wolf algorithm into the artificial bee colony algorithm,the GWOABC’s convergence speed and accuracy are improved.A new dynamic evaluation mechanism is introduced to improve the honey source selection and avoid the algorithm from getting trapped in local optima.Furthermore,the Cauchy mutation strategy is introduced in the later stage of the algorithm to further enhance its convergence speed.Simulation experiments show that GWOABC outperforms the compared algorithms in terms of convergence speed,convergence accuracy and robustness.Moreover,the path planned by GWOABC is superior and safer.(2)For the problem of UAV path planning in three-dimensional unknown environments,this thesis proposes a novel concept of saturation state and an improved Q-learning algorithm(Q-learning Algorithm Based On Saturation State,SQ-learning).SQ-learning abandons the traditional exploration-exploitation strategy and adopts a saturation-based exploration strategy.When a state is marked as saturated,no further exploration and exploitation is required,which saves a significant amount of training time.Thus,SQ-learning can dynamically adjust the planned trajectory to meet the needs of obstacle avoidance and path optimization when UAVs face unknown and variable environments.Simulation experiments show that the proposed saturation-based algorithm converges faster than traditional algorithms.(3)For the problem of multi-UAV path planning in three-dimensional unknown environments,this thesis proposes an improved Q-learning algorithm(Q-learning Algorithm Based On Pedestrian Motion Behavior,PQ-learning).By simulating human obstacle avoidance behaviors,the thesis introduces a reward function based on pedestrian motion models.This approach enables the trained UAVs to simulate pedestrian motion,not only avoiding obstacles and preventing collisions with other UAVs but also effectively reducing the total distance of multi-UAV path planning.Simulation results confirm the effectiveness of PQ-learning.