With a rise in the use of Unmanned Aerial Systems (UAS) across a variety of industries comes a rise in the concern for safety. Many available platforms nowadays come equipped with obstacle detection and collision avoidance systems. Systems such as these are often used for the creation of drone swarm platforms to better handle certain situations such as scanning a large area or increasing their coverage in a three-dimensional space. This paper explores the use of agent-based simulation to evaluate the characteristics and behaviors of collision avoidance for UAS swarms. The paper also explores the outcomes of allowing individual UAVs within a swarm to navigate semi-independently when an obstacle is detected and return to the swarm when a collision state is successfully prevented. Through the use of a scalable agent-based modeling approach, different swarm formation algorithms are modeled along with obstacles along the swarm’s flight path. Along with a view into the swarm’s collision avoidance behavior when an obstacle is encountered, unique differences can be seen in the characteristics of different formations. The formations modeled for the purposes of this paper include "follow the leader UAS", a horizontal “scan line”, a 3D cube, and a virtual forces formation. The performance characteristics that are investigated include the time to build the formation as well as the success rate for avoiding a collision. The model developed for this paper can be adapted to account for new formation algorithms as well as different collision avoidance algorithms in the future for finding an adequate solution for a given application.

Keywords
AGENT-BASED SIMULATION,MODELING,SIMULATIONS,UAV

Additional Keywords
UAS, Swarming