Collision avoidance among multiple aerial robots and other non-cooperative aircraft based on velocity planning

This paper presents a collision avoidance method for multiple UAV sharing the same aerial space with non- cooperative aircraft based on velocity planning. The proposed method finds a safe trajectory, modifying the velocity profile of the different vehicles involved in the collision and considering mobile obstacles. The proposed method applies two steps: Search Tree and Tabu Search. The objective is to find the nearest solution to the initially planned UAVs trajectories while meeting the time constraints on the execution of the algorithms. I. INTRODUCTION Multiple UAVs are cooperatively used to carry out tasks that can not be easily done by a single robot. In this context it appears the collision avoidance problem we address in this paper. The collision avoidance problem can be solved in two different ways. Collision free trajectories can be initially calculated before the vehicles start moving. This method has no significant computation time constraints. On the other hand, initial trajectories would be calculated without taking into account the trajectories of the other vehicles, and the potential collision would be solve in real time once they are detected. In this case computing time plays an important role. This is the problem we address in this paper. Recently the problem of motion planning for multiple robots has received a great deal of attention. In (12), the problem is written as a linear program subject to mixed integer constraints, known as a mixed-integer linear program (MILP). This can be solved using commercial software written by the Operation Research community. This problem has a significant complexity because of the high number of constraints and because it does not consider mobile obstacles. In (13) a method is proposed to geometrically construct a collision-free trajectory in (x,y,t) space. First, the authors evaluate the position and speed of the mobile obstacles. Assuming that the obstacles speed remain constant, they compute a set of oblique cylinders in (x,y,t) space to be avoided. The problem is then to find a trajectory connecting the initial position to a vertical line representing the goal. This method does not solve a multirobot motion planning problem, in which the trajectory of more than one robot can change to solve collisions. The method in (10) has the same drawback for the multi-robot problem considered in this paper. In this case a new trajectory of a vehicle, which