LPAC: Learnable Perception-Action-Communication Loops with Applications to Coverage Control
Saurav Agarwal, Ramya Muthukrishnan, Walker Gosrich, Vijay Kumar, Alejandro Ribeiro
AI summary
Problem
Decentralized robot swarms struggle to collaboratively monitor unknown environments due to limited sensing and communication constraints, causing traditional algorithms to scale poorly and miss critical observations.
Approach
LPAC combines a CNN for local perception, a GNN for decentralized message passing and aggregation, and an MLP for action generation, trained via imitation learning from a centralized planner.
Key results
- Outperforms standard decentralized and centralized coverage control baselines
- Generalizes to unseen feature distributions and varying swarm sizes
- Transfers to larger environments without performance degradation
- Remains robust to noisy position estimates and zero-shot on real-world data
Why it matters
Offers a scalable, communication-efficient framework for deploying robust multi-robot coordination in real-world applications like search-and-rescue and environmental monitoring.
Abstract
Coverage control is the problem of navigating a robot swarm to collaboratively monitor features or a phe- nomenon of interest not known a priori. The problem is chal- lenging in decentralized settings with robots that have limited communication and sensing capabilities. We propose a learn- able Perception-Action-Communication (LPAC) architecture for the problem, wherein a convolutional neural network (CNN) processes localized perception; a graph neural network (GNN) facilitates robot communications; finally, a shallow multi-layer perceptron (MLP) computes robot actions. The GNN enables collaboration in the robot swarm by computing what information to communicate with nearby robots and how to incorporate received information. Evaluations show that the LPAC models— trained using imitation learning—outperform standard decen- tralized and centralized coverage control algorithms. The learned policy generalizes to environments different from the training dataset, transfers to larger environments with more robots, and is robust to noisy position estimates. The results indicate the suitability of LPAC architectures for decentralized navigation in robot swarms to achieve collaborative behavior.