Artificial Intelligence1 min read

Robot swarms trained by trial and error in simulation keep working on real robots

By · Via Robohub · Published by Everything Blog

In short

The approach tackles the reality gap caused by friction, sensor noise, and mechanical differences by limiting the learning algorithm to coordinating predefined actions like exploration or light-seeking. During testing, these restricted setups significantly outperformed traditional state-of-the-art models that attempted direct wheel motor control in physical environments. The team demonstrated this efficacy across 250 real-world experiments utilizing a deployment of 20 physical e-puck robots. Such automated simulation methodologies could ultimately ready autonomous units for dangerous duties like environmental monitoring and post-earthquake search-and-rescue.

Key points

  • Simpler, restricted behavioral policies transfer to physical hardware more successfully t…: Simpler, restricted behavioral policies transfer to physical hardware more successfully than larger, unconstrained artificial intelligence models.
  • Restricting algorithms to predefined actions prevents robots from exploiting simulator qu…: Restricting algorithms to predefined actions prevents robots from exploiting simulator quirks instead of learning genuine cooperative behavior.
  • Future investigations will assess the training technique across broader swarms, novel rob…: Future investigations will assess the training technique across broader swarms, novel robotic platforms, and less controlled operational environments.

Sources checked by our analysis pipeline: 1

Reinforcement learning lets robots learn behaviors on their own, through trial and error, usually inside a computer simulation. Moving what they learn onto physical robots is another matter. Simulated and real robots never behave quite the same way, a mismatch known as the “reality gap,” and reinforcement learning has long been particularly sensitive to it. Researchers at IRIDIA, the artificial intelligence laboratory of the Université libre de Bruxelles (ULB), have shown how to get around it for robot swarms: swarms trained with their method in simulation keep performing reliably once they are moved onto real robots.

Teaching a swarm of robots to search for survivors after an earthquake, or to explore a hostile site, entirely inside a computer before risking a real machine sounds like the sensible way to do it. The catch is the reality gap. Friction, sensor noise and mechanical imprecision rarely behave the way they do in a simulation, and skills learned there tend to fall apart when a robot meets the real world.

Ilyes Gharbi, Mauro Birattari and colleagues at IRIDIA (Brussels School of Engineering, ULB)…

Artificial Intelligence