Why do animals use certain sensorimotor control strategies instead of others? To explore this question, we’ve developed a large-scale swarm robotics platform—including aerial drones, land-based robots, and underwater robots—that allows us to test biologically inspired control systems in real-world conditions.

By implementing these natural control models in robots and comparing them with engineering-optimized strategies such as optimal control, model predictive control (MPC), and reinforcement learning based control, we can evaluate their performance across a range of tasks. This comparison helps us uncover the advantages and trade-offs of evolutionarily shaped behaviors.

One of the key innovations in this project is the development of a vision-based swarm system, which enables coordination among many robots without relying on centralized communication. This approach allows for scalable experiments that mirror the distributed nature of collective behavior in animals.

Through this reciprocal research, termed “RoboTwin”—where biology inspires robotics and robotics helps test biological hypotheses—we aim to better understand both natural intelligence and how to design more robust, adaptive robotic systems.

Back to All Research