Problem
Research questions and friction points this paper is trying to address.
Robotics
Self-organization
Adaptive Systems
Collective Intelligence for 2D Push Manipulations With Mobile Robots
🤖 AI Summary
This work addresses the challenge of collaborative object pushing by multiple robots in 2D environments, where existing approaches suffer from weak self-organization, poor environmental adaptability, and insufficient task dynamism. We propose a neural distillation framework grounded in soft-body physics simulation: an optimization-based planner is first trained in simulation, and its decision-making knowledge is distilled into a lightweight attention-based neural network, unifying generalization capability with online adaptability. Crucially, the method avoids explicit physical parameter modeling, enabling zero-shot transfer to unseen robot configurations and robust real-time response to environmental disturbances. Experimental results demonstrate stable task completion across diverse unseen robot formations, significantly outperforming conventional multi-agent reinforcement learning and model predictive control baselines in both success rate and adaptability.
📝 Abstract
While natural systems often present collective intelligence that allows them to self-organize and adapt to changes, the equivalent is missing in most artificial systems. We explore the possibility of such a system in the context of cooperative 2D push manipulations using mobile robots. Although conventional works demonstrate potential solutions for the problem in restricted settings, they have computational and learning difficulties. More importantly, these systems do not possess the ability to adapt when facing environmental changes. In this work, we show that by distilling a planner derived from a differentiable soft-body physics simulator into an attention-based neural network, our multi-robot push manipulation system achieves better performance than baselines. In addition, our system also generalizes to configurations not seen during training and is able to adapt toward task completions when external turbulence and environmental changes are applied. Supplementary videos can be found on our project website: https://sites.google.com/view/ciom/home.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Attention-based Neural Networks
Adaptive Collaboration
Self-organization in Robotics
🔎 Similar Papers
No similar papers found.
