Development of a predictive search model of flocking for robot swarms

Önür, Giray
One of the main challenges in swarm robotics is achieving robust and scalable flocking, allowing large numbers of robots to move together in a coordinated and cohesive manner while avoiding obstacles or threats. Flocking models in swarm robotic systems typically employ reactive behaviors, such as cohesion, alignment, and avoidance. The use of potential fields has enabled the derivation of reactive control laws by utilizing obstacles and neighboring robots as sources of force for flocking. However, reactive behaviors, especially when multiple of them are simultaneously active, as in the case of flocking, can lead to collisions or inefficient motion within the flock due to their short-sighted approach. Approaches aiming to generate smoother and optimal flocking, such as the use of Model Predictive Control, either require centralized coordination or distributed coordination, which necessitates low-latency and high-bandwidth communication within the swarm, as well as substantial computational resources. In this thesis, a predictive search model which generates smoother and safer flocking of robotic swarms in the presence of obstacles while efficiently considering the predicted states of other robots is introduced. The proposed model is evaluated in simulated environments with both static and dynamic obstacles, and its performance is compared with a potential field flocking model. Furthermore, the performance of the predictive search model is validated in a real-world scenario involving a swarm of six indoor quadcopters navigating static and dynamic obstacles. The results demonstrate that the predictive search model produces smoother and safer flocking in swarm robotic systems with limited sensing capabilities and computational resources.
Citation Formats
G. Önür, “Development of a predictive search model of flocking for robot swarms,” M.S. - Master of Science, Middle East Technical University, 2023.