Self-assembling robots
Such an artefact called swarm-bot system is composed of a number of simpler, insect-like robots, the s-bots. The activity where more than one s-bots physically connect to each other to form a swarm-bot constitutes one of the most distinguishing characteristics of the swarm-bot system, the autonomous self-assembling. The key idea behind this relies on the fact that difficult situations can not cannot be addressed by a single s-bot requiring the cooperative effort performed by the swarm-bot system. Therefore, self-assembling can improve the efficiency of an aggregation of autonomous cooperating s-bots under various complex environmental conditions. For instance, self-assembling reduces the risk of toppling over when moving in a particularly rough terrain. Self-assembling could also be useful for bridging a gap between the two sides of a trough that is larger in size than an s-bot, minimising the possibilities of falling in it. A swarm-bot system could also push/pull a massive or heavy object more easily than a single s-bot. Self-assembling is initiated with an s-bot turning its red lights on, attracting s-bots being in within its proximity by perceiving the red lights. The s-bots approach the red-lit s-bot and connect to it by grasping the red s-bot ring with their gripper. After a successful connection, they turn on their lights to attract other s-bots. In case of difficulties in approaching, there is a backward movement followed by a new attempt to re-approach. The developed self-assembling controllers have been based on a perceptron-type neural network whose weights evolved using an evolutionary algorithm. Synthesis of these controllers was performed during simulation of up to five simulated s-bots and then ported to the real s-bots. Experimental results showed that the procedure is reliable in controlling the connection of s-bots to each other or to special luminous objects (s-toys). Additionally, the procedure features scalability as up to 16 s-bots have successfully undergone experimental tests. Most importantly, it is a robust procedure exerting control to self-assembling s-bots moving on both flat and moderately rough terrain.
