Uploaded May 2025 | Updated September 2026, 1 hour ago
Future Leaders in Robotics and AI Seminar Series: Provably Scalable and Decentralized Control Design for Minimalist Robot Swarms
Online Seminar
Himani Sinhmar
Postdoctoral Fellow
Princeton University
How much collective intelligence can emerge from individually simple agents? In this talk, I explore this question through the lens of minimalist robotic swarms—systems composed of robots operating without communication, GPS, or global reference frames, using only minimal sensing and lightweight onboard computation. I present a control-theoretic framework inspired by opinion dynamics—originally studied in social systems— adapted here for real-time robotic decision-making. By evolving internal dynamical states based solely on local observations, agents achieve emergent behaviors such as consensus for coordination and dissensus for resolving conflicts, including deadlock, in dynamic, partially observable environments. A core contribution of this work is the mathematical characterization of performance guarantees and design trade-offs: for example, how many agents are needed to offset limited sensing? How do sensing constraints impact coordination speed and safety? What minimal assumptions ensure correctness? These questions define a principled design space linking theoretical guarantees to practical considerations such as cost, robustness, and scalability. Beyond navigation, I extend this framework to distributed resource allocation, showing how multi-option opinion dynamics enable decentralized task selection under uncertainty, paving the way toward robust, learning-free swarm algorithms. If you're interested in how minimalist principles can yield scalable, reliable swarm behaviors without relying on centralized control or intensive computation, this talk will offer both the theoretical foundations and practical tools to get you there.
For more information, please visit:
https://robotics.umd.edu/FutureLeaders
Future Leaders in Robotics and AI Seminar Series: Provably Scalable and Decentralized Control Design for Minimalist Robot Swarms
Online Seminar
Himani Sinhmar
Postdoctoral Fellow
Princeton University
How much collective intelligence can emerge from individually simple agents? In this talk, I explore this question through the lens of minimalist robotic swarms—systems composed of robots operating without communication, GPS, or global reference frames, using only minimal sensing and lightweight onboard computation. I present a control-theoretic framework inspired by opinion dynamics—originally studied in social systems— adapted here for real-time robotic decision-making. By evolving internal dynamical states based solely on local observations, agents achieve emergent behaviors such as consensus for coordination and dissensus for resolving conflicts, including deadlock, in dynamic, partially observable environments. A core contribution of this work is the mathematical characterization of performance guarantees and design trade-offs: for example, how many agents are needed to offset limited sensing? How do sensing constraints impact coordination speed and safety? What minimal assumptions ensure correctness? These questions define a principled design space linking theoretical guarantees to practical considerations such as cost, robustness, and scalability. Beyond navigation, I extend this framework to distributed resource allocation, showing how multi-option opinion dynamics enable decentralized task selection under uncertainty, paving the way toward robust, learning-free swarm algorithms. If you're interested in how minimalist principles can yield scalable, reliable swarm behaviors without relying on centralized control or intensive computation, this talk will offer both the theoretical foundations and practical tools to get you there.
For more information, please visit:
https://robotics.umd.edu/FutureLeaders
