Abstract:
In multi-robot systems, maintaining joint action consistency is crucial, especially when individual robots have inconsistent beliefs about the world. This thesis explores two methods to address this challenge. The first method utilizes factor graphs to model dependencies between robots’ actions and beliefs, providing a structured way to handle inconsistencies. The second method applies Bayesian games to optimize actions by modeling the interaction between robots as a game where each robot maximizes its utility considering the beliefs of others. These methods aim to improve the robustness and coordination of multi-robot systems in uncertain environments, leading to more efficient collaboration and decision-making in the presence of belief discrepancies.
