Event
Reinforcement learning enables lunar robot swarms to spontaneously huddle for warmth
1 reports · 1 sources
Anais Jeger and co-authors report in a paper that training a single policy shared by twenty lunar robots with reinforcement learning, using only the average temperature of the five coldest robots at the end of the night as a one-time reward, led to the emergence of two cooperative heat-conservation behaviors: clustering together and rotating between the core and the edge. They measured a minimum stored energy of 155.6 +/- 1.3 pcm-h per robot required for survival (1 pcm-h is 7.2 kJ of stored heat), 37% lower than stationary huddling and within 13% of a physics-based lower bound; under an energy budget in which stationary huddling would lose six robots, all robots survived. The behavior requires neither a shared spatial reference nor robots touching one another, and transfers directly from twenty to forty robots without retraining. The paper has been submitted to IEEE Robotics and Automation Letters; the above results are as reported by the authors.
Report timeline
- Lunar swarm robots exhibit emergent cooperative thermoregulation via reinforcement learning
The study used reinforcement learning to train a single policy shared by a set of twenty robots, with only the mean temperature of the five coldest robots at the end of the night as a terminal reward, and two behaviors emerged: aggregation and rotation between the core and the edge. The minimum stored energy required per robot for survival was measured at 155.6 +/- 1.3 pcm-h (1 pcm-h is 7.2 kJ of stored heat), 37% lower than stationary huddling and within 13% of a physics-based lower bound; under an energy budget in which stationary huddling would lose six robots, all robots survived.
arXiv cs.RO ·