Adaptivity and self-repair in robot self-assembly

Abstract

Self-assembly is a natural process of autonomously forming structures from a collection of simple components. In swarm robotics, it is an open challenge to self-assemble scalable and robust structures that can adapt to dynamic features of the environment. We take a photomorphogenetic approach—a method directed by light-stimuli for multi-robot self-assembly inspired by the tissue growth of trees—and a honeybee-inspired model. Existing research on multi-robot self-assembly is mostly limited to predefined shapes that reconfigure only on long time-scales. Here the state-of-the-art is extended, as the swarm autonomously rearranges the assembled structure to react to dynamic environments and repair damage. The high turnover rate of adding robots to the structure and allowing them to leave again creates novel challenges of how to ensure minimal stability as well as how to balance exploration and exploitation of the assembly. An adaptive resource distribution method similar to a plant’s vascular system steers the assembly process. Robots aggregate into a tree structure and receive virtual resource according to local environmental features—here, specifically light. The effectiveness of our approach is validated through several real and simulated robot experiments consisting of five components. (1) Leader selection: during the first set of experiments the robot swarm collectively selects a leader and a place to initiate self-assembly. The robots are exposed to a gradient of light that is bright on one side and gradually dimming to the other. The task is to initiate a tree structure in the darkest area that is implemented by a honeybee-inspired approach. (2) Directed aggregation: a directed aggregation in the form of a tree structure grows towards the light source. (3) Adaptation to dynamic environment: an improvement is then to create structures that adapt to the environment not only during the formation process but also continuously throughout the experiments. Robots in the dark areas fail to absorb enough resource to keep them in the structure, while the aggregation grows in areas of higher quality. The swarm adapts to the dynamic light setup by continuously allocating the resource to the part of the structure in the brighter area. (4) Site selection: we take one step further to test the robots’ ability to adapt to changes and to collectively select the most advantageous growth site in the arena based on the brightness and the proximity of the sites. The swarm succeeds in finding and selecting the more advantageous site and succeeds in adapting its choice after changes in the environment. (5) Self-repair: we evaluate the robustness of our method by testing the swarm’s ability to regrow damaged areas. Soon after the damage, the tree structure grows back, repairing the structure. Simulation of a swarm of 1024 robots demonstrates the scalability of our adaptive self-assembly method. The thesis therefore contributes to a broadened foundation for stimuli-driven self-assembly that is adaptive and robust. As in many works on robot self-assembly, we also face the problem of finding and defining the appropriate hardware approach and future work has to prove that we can govern the hardware challenges.

More information

Identifiers

ORCID for Mohammad Soorati: orcid.org/0000-0001-6954-1284

Catalogue record

Export record