Research On Obstacle Avoidance For Multi-Rotor MAVs Based On Layered A* Algorithm

In recent years,multi-rotor MAV(Micro Air Vehicles)technology has developed rapidly,and the MAV has been used in military and civil applications.Micro Aerial Vehicle(MAV)is an ideal platform for performing autonomous missions in 3D space and has received intense attention.A basic requirement for performing such task is obstacle avoidance,where the MAV detects obstacles,and plans a path to avoid collision.Different from outdoor environments,avoiding various obstacles densely distributed indoors is a more challenging topic.We focused on obstacle avoidance for MAVs in indoor complex environments,proposed a more effective layered A* path planning algorithm in 3D space.The basic principle and recent researches of obstacle avoidance for MAVs are introduced at first.Then details of the stereo vision,A* static path planning algorithm and D* dynamic path planning algorithm are produced as theoretical basis.After that the algorithm of the proposed layered A* path planning is presented in detail.By fusing depth from both the stereo camera and the ultrasonic sensor,our system improves the capability of detection for objects that are difficult to be detected by vision alone,and reduces detection outliers that causes false alarms.In order to run in real time on the ARM processor board,a layered A* path planning algorithm is proposed that leads our system to faster path planning in 3D pace.Finally,experiments in actual indoor environment are conducted to verify the availability and robustness in experiments.Three groups of experiments are included,and they are obstacle detection experiments,obstacle avoidance in static environments and dynamic environments respectively.On ARM test board,the processing time per frame of proposed layered A* algorithm is 0.12 s,better than 0.48 s for A* static algorithm and D* dynamic algorithm.