Spring 2024 GRASP Seminar: Himani Sinhmar, Cornell University, “Control Design With Guarantees for Minimalist Multi-Agent Robotic Systems”

This is a hybrid event with a VIRTUAL SPEAKER. There will be in-person attendance in Raisler Lounge and virtual attendance on Zoom.

ABSTRACT

While there has been significant progress in expanding the capabilities of multi-agent systems, a key challenge is the coordination and navigation of robot teams, especially in unknown, unstructured, and GPS-denied environments with limited communication. This paradigm raises intriguing questions: given a task, how minimal can an agent’s abilities be to guarantee the desired behavior while ensuring safety and robustness? Can we design a framework that adapts to an individual agent’s limitations, adjusting behaviors based on the task requirements, memory usage, or sensory capabilities? What quantifiable trade-offs emerge when balancing onboard capabilities with the control design for minimal robots? In this talk, I will delve deeper into these compelling questions, focusing on guaranteeing the globally desired behavior of a team of minimalistic robots. I will show how we can design provably correct reactive controllers that rely solely on a robot’s simple sensors to construct an impression of its surroundings for a multi-agent target search and encapsulation in a dynamic, unknown environment. The framework provides several quantifiable trade-offs between task, system, and control design parameters. I will also provide theoretical guarantees for encapsulating faster evaders where each agent lacks knowledge of the evader’s motion, relaxing a strong assumption typically required in pursuit-evasion literature.