Generative adversarial imitation learning for robot swarms: Learning from human demonstrations and trained policies
This work addresses the challenge of enabling robot swarms to learn complex collective tasks from demonstrations, which is incremental as it adapts existing imitation learning methods to swarm robotics with real-world validation.
The paper tackled the problem of teaching robot swarms collective behaviors by introducing a generative adversarial imitation learning framework that learns from both human demonstrations and existing policy rollouts. Results showed the framework successfully learned meaningful behaviors across six missions, achieving performance comparable to the demonstrations in simulation and preserving behavior characteristics when deployed on real TurtleBot 4 robots.
In imitation learning, robots are supposed to learn from demonstrations of the desired behavior. Most of the work in imitation learning for swarm robotics provides the demonstrations as rollouts of an existing policy. In this work, we provide a framework based on generative adversarial imitation learning that aims to learn collective behaviors from human demonstrations. Our framework is evaluated across six different missions, learning both from manual demonstrations and demonstrations derived from a PPO-trained policy. Results show that the imitation learning process is able to learn qualitatively meaningful behaviors that perform similarly well as the provided demonstrations. Additionally, we deploy the learned policies on a swarm of TurtleBot 4 robots in real-robot experiments. The exhibited behaviors preserved their visually recognizable character and their performance is comparable to the one achieved in simulation.