Model predictive control for obstacle avoidance in drone swarms

Abstract

In recent decades, Unmanned Aerial Vehicles (UAVs) have seen increasingly many applications and configurations with the advancement of UAV technology. The term “drones” may also be interchangeably used and associated with these UAVs. One application of these drones is in the form of a drone swarm, where multiple drones fly together to accomplish certain tasks. The number and size of the drones used in the swarm may vary depending on the desired scale of the swarm and each drone may either be controlled in a centralised or decentralised manner. However, an issue that is present in drone swarms is the issue of obstacle avoidance. Each drone in the swarm must now avoid the other drones in its vicinity which leads to a greater number of obstacles to avoid in its environment as compared to the situation in which there is no drone swarm. Model Predictive Control (MPC) presents potential to resolve this issue by transforming the issue of obstacle avoidance into an optimisation problem. By attempting to optimise a cost function and treating the possibility of collision as an additional cost to be considered, it increases the chance of the drone avoiding obstacles. In this project, the concepts of MPC were used to simulate a drone avoiding an obstacle. If the drone detected an obstacle in front of it, it would follow a certain trajectory until the obstacle was no longer directly in front of it. This simulation was done as part of the author and his team’s preparation for the annual Singapore Amazing Flying Machine Competition (SAFMC) 2024. In this competition, a swarm of drones was tasked to navigate a search area and land at specific locations. Due to the large number of drones and the small space between each drone, collision between drones was an issue. Thus, it was thought that using the concepts of MPC to control the drones to follow a certain trajectory while waiting for the other drones in front of it to fly past, thereby clearing the way, would serve as an appropriate form of obstacle avoidance. However, due to hardware limitations of the drone used, concessions for obstacle avoidance had to be made. Nonetheless, the author and his team competed valiantly against many experienced competitors and emerged fourth in the competition.