Using Game Theory to Improve Robot Team Collaboration and Coordination
Source: https://www.sciencedirect.com/science/article/pii/S1474667016417763
In a research study conducted by the University of Sheffield, the problem of coordinating actions in robot teams to accomplish tasks in an efficient manner was tackled by using game theory learning algorithms. According to the paper, the novel cooperative control methodology they implemented to address the issue was rooted in the idea that each robot predicts the actions of others to make a decision for the completion of the task while maximizing their own reward. In particular, robots in collaborative teams aim to reduce their cost function by interacting and communicating with other robots. In their methodology, robots learn each others policies and employ this information to educate their actions in order to minimize their cost functions.
Researchers tested their algorithm in a warehouse setting for collaborative tasks between material handling robots and patrolling robots. The study’s results indicate that the learning algorithm used converges to the Nash equilibrium of a given game. The research team hopes to further enhance their proposed controller design methodology to coordination between teams of ground-based and aerial robots. By applying game theory to robot team collaboration and coordination scenarios, one can make automated task completion more effective, thereby simplifying how humans go about executing various functionalities and satisfying needs in different applications of robots and artificial intelligence. Understanding the interdependence of actions in collaborative robot teams is crucial in order to achieve this goal so that researchers can maximize efficiency in task execution. In doing so, higher performance metrics can be derived from robot teams.
The foundation upon which researchers of this study based the development of their learning algorithms directly relates to the ideas covered in class about game theory and rational decision making. This study is a present-day application of how game theory can be used in cutting-edge research in the fields of robotics, computer science, and systems engineering. Game theory was used in this study to understand and maximize positive decision-making between robots that would lead to better performance in collaborative tasks. The results detailed in this paper parallel the ideas and outcomes uncovered in class in regards to how mathematical models and strategy can be used to maximize payoffs of individual players, which in this case are autonomous agents.