Modelling and simulating obstacle avoidance for a UAV

Abstract

UAV technology is frequently used in various fields of industry like aerial photography, agricultural irrigation, rescue after accidents and unmanned logistics. In whichever situation, obstacle avoidance is the most important and difficult part for engineers. There have been substantial researches on UAV obstacle avoidance, which can be classified as traditional methods and learning-based method. In industrial application, traditional methods usually ensure the lower limit of performance, so my dissertation is making improvements of the traditional planning and control methods and test their viability and effectiveness. In path planning section, I design improved A* search algorithm, which considers the influence on searching brought by UAV body and decreases calculation consumption by skipping unnecessary searches. I also design hybrid A* search algorithm using Dubins Curve as heuristic function to increase smoothness of path. For trajectory optimization, I applied gradient descent method and QP optimization method. For the two ways, I both design suitable and reasonable objective functions and optimize the trajectory in the aspects of both smoothness and obstacle avoidance. In order to track the reference trajectory well, I decouple the whole dynamic system and design cascade PID controllers, which includes position tracking and attitude tracking. On the contrary, I design MPC controller for the whole system, which simultaneously tracks reference positions and angles. Particularly, I do the linearization to the state space equation and amendment to the objective function so as to adjust to the complicated nonlinear dynamic system and reduce calculation complexity. Through simulation, we can find that the proposed methods above can generate smooth reference trajectories and make UAV perform well in tracking them.