MEAM Seminar: “Macroscopic Ensemble Methods for Robot Team Data Collection in Dynamic Environments”

Data is necessary to improve our understanding of dynamic environments, like the ocean. With limited sensing resources, the challenge lies in identifying and acquiring sensor measurements over large spatial and temporal scales. One solution is to use a team of robots equipped with sensors to collect data. However, robot teams still require methods that effectively tell the robots where to sample informative data. Existing approaches develop plans for each individual robot which works well if the team is small (less than 10 agents) and operating in a simple environment (an open field). Unfortunately, these solutions require careful engineering and cannot be easily adapted to changing environmental conditions. For this reason, we want to model team-wide objectives using dynamical systems theory. Specifically, our robot team modeling technique is called macroscopic ensemble modeling. These methods are known to easily control large robot teams (more than 50 robots) and even scale to control many different types of robots. Nevertheless, macroscopic ensemble methods require extensions to effectively distribute robots in dynamic environments. This seminar will cover our recent results incorporating both collaborations and environmental feedback into macroscopic ensemble robot team models. Our results demonstrate novel team-wide behavior beneficial to collecting data in dynamic environments.