Parallel Task Execution, Morphology Control and Scalability in a Swarm of Self-Assembling Robots

Abstract —We investigate the scalability of a morpholog-ically flexible self-assembling robotic system by measuringtask execution performance. We use a scenario consisting ofthree subtasks — gap crossing, bridge traversal and objectpushing. Each subtask can only be solved by a dedicatedself-assembled morphology. To successfully complete the sce-nario, individual robots must autonomously assemble anddisassemble to form morphologies appropriate to the subtaskat hand. Environmental cues tell the robots when they haveencountered a particular task. Parallel execution of tasksis possible when there is a sufficient number of robots.With simulated robots, we perform a series of experimentsdemonstrating the feasibility and the scalability of our system.We implement our distributed control using the scripting lan-guage SWARMORPH-script that has been used in previousstudies to form morphologies with up to nine real robots. I. I NTRODUCTION Self-assembling robotic systems are composed of multi-ple autonomous agents that can physically connect to eachother to form larger composite robotic entities. Two of thekey potential benefits of self-assembling robotic systemsare morphological flexibility and parallelism. Morphologi-cal flexibility is important because any robotic entity musthave a morphology that is in some way appropriate to thetask it needs to perform. In theory, the ability to form awide range of different morphologies should allow futureself-assembling systems to tackle a wider range of tasksthan conventional monolithic robots. Such self-assemblingsystems may well comprise thousands or even millions ofindividual agents. In such large systems, parallelism willbe the key to efficiency—different self-assembled roboticentities will be able to carry out different tasks at the sametime. A well-designed self-assembling system should thusallow for massively parallel task execution.In this study, we explore a scenario designed to investi-gate morphological flexibility and large scale parallelism.In our scenario, a series of subtasks must be completed.Each subtask is solvable by a dedicated self-assembledmorphology, which is incapable of solving the other sub-tasks. The robots start at one end of the arena and performphototaxis towards a light source at the other end of thearena. As they proceed, environmental cues indicate thepresence of particular subtasks to be solved. When theyencounter a subtask, the robots must assemble into theappropriate morphology for the subtask at hand. Once thatsubtask is complete, the robots disassemble and continuephototaxis. They are thus ready to assemble into anothermorphology as soon as they encounter another subtask.The nature of the subtasks allows for a degree of parallelexecution.