Kinetic theory of decentralized learning for smart active matter
arXiv:2501.03948 · doi:10.1103/5m44-kwhv
Abstract
Smart active matter has the ability to control its motion guided by individual policies to achieve collective goals. We introduce a theoretical framework to study a decentralized learning process in which agents can locally exchange policies to adapt their behavior and maximize a predefined reward function. We use our formalism to derive explicit hydrodynamic equations for the policy dynamics. We apply the theory to two different microscopic models where policies correspond either to fixed parameters similar to evolutionary dynamics, or to state-dependent controllers known from the field of robotics. We find good agreement between theoretical predictions and agent-based simulations. By deriving fundamental control parameters and uncertainty relations, our work lays the foundations for a statistical physics analysis of decentralized learning.
Phys. Rev. Lett. 134, 248302 (Editors' Suggestion)
References in corpus (28)
- Statistical physics of social dynamics
- The Mechanics and Statistics of Active Matter
- Active Brownian Particles. From Individual to Collective Stochastic Dynamics
- Continuous Opinion Dynamics under Bounded Confidence: A Survey
- Boltzmann and hydrodynamic description for self-propelled particles
- Biological Evolution and Statistical Physics
- Hydrodynamic equations for self-propelled particles: microscopic derivation and stability analysis
- Flow Navigation by Smart Microswimmers via Reinforcement Learning
- Kinetic Theory of Flocking: Derivation of Hydrodynamic Equations
- Mesoscopic theory for fluctuating active nematics
- Embodied Evolution in Collective Robotics: A Review
- Demonstration of Decentralized, Physics-Driven Learning
- Supervised learning in a mechanical system
- Morphological computation and decentralized learning in a swarm of sterically interacting robots
- Learning to flock through reinforcement
- Programming Active Cohesive Granular Matter with Mechanically Induced Phase Changes
- Learning to Control Active Matter
- Photofocusing: Light and flow of phototactic microswimmer suspension
- Optimal active particle navigation meets machine learning
- Markovian robots: minimal navigation strategies for active particles
- Towards a quantitative kinetic theory of polar active matter
- Learned multi-stability in mechanical networks
- Optimal navigation of microswimmers in complex and noisy environments
- Optimal navigation strategies for microswimmers on curved manifolds
- Body and mind: Decoding the dynamics of pedestrians and the effect of smartphone distraction by coupling mechanical and decisional processes
- Statistical Mechanics of Horizontal Gene Transfer in Evolutionary Ecology
- Optimal collision avoidance in swarms of active Brownian particles
- Learning to swim efficiently in a nonuniform flow field