A distributed and parallel self-assembly approach for swarm robotics

This paper presents a novel distributed and parallel self-assembly approach, which uses the lattice system as a systematic structure and homogeneous robots as shaping carriers to form a two-dimensional user-specified shape autonomously. Given a desired shape to be formed, the initial shape matches with it to execute the initialization of all individuals to allow each of them gets its location and status. Based on that, with the stratified mechanism, the macro-level behavior of large-scale group formation in swarm robotics is transformed to local formation action of individuals within the edge layers of the current aggregate, which makes complex shape formation possible. Then two motion-chains, a collection of individuals that have priority to move currently, are autonomously planning in parallel through local interaction and collaboration. Once all robots within a motion-chain are activated, each of them will move along the outer edge of the current aggregate orderly to fill the edge-filling layer efficiently. The motion-chains will be iteratively generated until the desired shape is formed. We evaluate the feasibility and scalability of this novel approach in simulation-based experiments, and implement the self-assembly algorithm on the Rubik robot, a hardware system developed in our lab.