제목: Desnity-Driven Optimal Control of an Autonomous Multi-Agent System with an Application to Planetary Explorations​​​​​​​​​​​​​​​

장소: 과학관 B101호 

Abstract:

A multi-agent system (or multi-robot system) has great potential to be applied to wide applications as it can provide many advantages over a single agent such as enhanced efficiency and adaptability while being less prone to a single point of failure. With an increased interest in the multi-agent system, NASA has been developing a team of rovers for the CADRE (Cooperative Autonomous Distributed Robotic Exploration) project to develop a network of small rovers. These rovers will work together to explore the Moon with autonomy, making decisions and acting without the need for constant human intervention. In this mission, exploration efficiency will likely be the key element because the size of the domain (or quantity of interest) is very huge whereas the number of agents, their energies, and communication range are finite, necessitating a transformative idea to cover areas of interest efficiently rather than uniformly. As a result, there is a need to develop a new multi-agent control algorithm to maximize multi-agent coverage efficiency while providing a holistic approach for collaboration between multiple agents.

This seminar introduces a density-driven multi-agent exploration scheme through the Optimal Transport theory for such efficiency. This work is based on our research group’s recent works supported by NSF and NASA to develop a new multi-agent control algorithm. By taking into account a probability density function (PDF) given as reference density, this scheme will bring a novel idea to multi-agent coverage problems.