Math Biology Seminar
Large-scale coordination tasks are well executed in nature by animal groups exhibiting collective behavior, such as fish schools and bird flocks, where complex structures emerge from decisions based on local information without a centralized leader. Among these groups, bat swarms stand out as a unique example, since they can perform sophisticated active sensing (echolocation) for navigation and hunting which is intercepted and may be used by their peers. Unlike groups that only use passive sensing such as vision, multi-agent systems inspired by bat swarms may be able to achieve types of collective behavior applicable to engineered systems using active sensing, such as robotic teams using sonar or lidar. In this talk, we first explore model-free techniques that can be used to uncover interactions between bats flying together in data from a wild bat colony. We then write an agent-based models that generates collective behavior using passive and active sensing. Finally, we explore how such multi-modal sensing may be achieved for individual systems tackling a canonical robotics problem, simultaneous localization and mapping or SLAM.