Research On UAV Collision Avoidance Decision In Urban Low-altitude Environment

For decision-making problem of UAV collision avoidance in low-altitude urban complex terrain environment and dense traffic flow,to model and describe this problem through the Markov decision process.Due to the commonly used online solution algorithm in the Markov decision process—Monte carlo tree search algorithm,the search depth is limited to ensure the real-time performance of the algorithm,which makes the algorithm easy to fall into a local optimal solution.Due to the jump point search algorithm has the characteristics of rapid planning of the global optimum,this paper establishes discrete waypoints to guide the UAV by jumping point search in the UAV operating space and improving the reward function to weigh the flight path,finally realizes the global collision avoidance against static obstacles while performing dynamic collision avoidance.Firstly,model the space in which the UAV operates,and express the real environmental information as a mathematical theoretical model;and express the UAV motion model in the simulation as an aircraft with six degrees of freedom.Secondly,according to the Markov decision process theory,the flight collision process of the UAV is discretized into a state set composed of different state vectors;discrete the executable actions of the drone into a set of actions consisting of a finite number of flight actions;improve the reward evaluation function to make it possible to weigh the choice of the UAV optimal route and dynamic collision avoidance.Aiming at the simulation verification of the two-dimensional collision avoidance problem of UAVs at the same altitude,this paper establishes three typical simulation scenarios to simulate the UAV’s collision avoidance effect on dynamic targets in the restricted airspace,and uses visual simulation to verify The algorithm has advantages over only the Monte Carlo tree search algorithm.For obstacles formed by continuous high-rise buildings,the UAV will fly around in a twodimensional plane to avoid it,which will not only make the flight path longer,but also consume more energy and reduce the endurance time.Therefore,based on the two-dimensional collision avoidance algorithm,the action set is expanded by adding altitude change actions,and the UAV collision avoidance algorithm is expanded from a two-dimensional plane to a three-dimensional space,so that the UAV can take a climb action to change the flight altitude The collision avoidance strategy overtakes obstacles,and then changes to level flight or descends to the previous flight height according to the actual situation.Finally,it is verified in the simulation model,which shows that the three-dimensional collision avoidance algorithm can reasonably choose the strategy for overcoming obstacles and flying around.