Swarm-Inspired Controller: An Inference-Free Approach to Distributed Manipulation

Nicolas Bessone, Kasper Støy, Payam Zahadat

Abstract: This paper proposes an approach to a self-organizing decentralized controller for modular robotic manipulation surfaces. Distributed manipulator surfaces use a grid of independently controlled actuators to induce motion on objects through multiple contact points. These systems are primarily designed to achieve precise positioning and orientation of objects. The proposed approach is introduced in the context of a simulated virtual environment featuring a generic distributed planar manipulator. Due to its high level of abstraction, this approach is well-suited for use with different actuation principles and sensing devices, concentrating its focus on the information and module capabilities required to perform the task. A series of experiments were designed to assess the system’s self-organization capabilities for objects of different shapes alongside assessing its performance and fault tolerance. We compared the results of the proposed inference-free method against a technique that relies on information about global object properties, where this information is provided by an external source or inferred by the actuators, finding comparable outcomes. The experiments provide evidence of the effectiveness of the developed control rules, thereby highlighting the approach’s potential as a viable solution for achieving efficient and resilient decentralized control for robotic manipulation surfaces.