Evolutionary Distributed Control of a Biologically Inspired Modular Robot

Arguably the innovative problem solving abilities is one of the cornerstones for ensuring the survival of the homo sapien species in the game of evolution. Throughout history, when faced with challenges it was not uncommon for mankind to turn to nature for answers. In the modern day, problem solving utilizing techniques harnessed from nature has become a niche of the computational intelligence field. There are quite a number of classical contributions in this respect, which include artificial neural networks (ANN), genetic algorithm (GA), ant colony optimization (ACO), cellular automata (CA) and artificial immune system (AIS). In the spirit of drawing inspiration from nature, our laboratory developed a modular robot modelled after a marine dwelling organism called the brittle star. The robot consists of independent modules with each module incorporating an onboard microcontroller for governing the behaviour of the module, actuator for inducing motion, and touch sensors for feeling the environment. Robot of such nature can be useful in search and rescue operations; for instance during earthquake the robot can be deployed to seek for survivors trapped under collapsed buildings which would otherwise be hazardous for human rescuers to reach. Before novel applications of the robot can be envisioned the fundamental issues of motion control needs to be addressed. While the notion of studying the motion characteristics of the brittle star and incorporating it into the robot is intuitive and insightful, it is nonetheless quite impractical. The reason for this stems from the fact that the range of motion of the highly agile arms of brittle star in an aqueous environment overwhelmingly surpasses the two-degree of freedom legs of the landlocked robot. Hence we turn to nature once again, and attempt to draw from the evolutionary phenomena as a driving force to evolve the robot to move in its environment pretty much the same way the biological organisms have been shaped over billions of years by adapting to their environments. Evolutionary algorithms are computational models, which capture the essence of evolution. Developed by John Holland in early 1970s (Holland, 1975), genetic algorithm (GA) is one of the widely adopted evolutionary algorithms. Inspired by biological adaptations, genetic algorithm is essentially a search technique used extensively to solve optimization related problems (Goldberg, 1989). The algorithm involves representation of candidate solutions to a problem using chromosomes also known as individuals. The initial randomly generated population of individuals is successively transformed based on their fitness by applying genetic operators