Using Evolution to Design Modular Robots: An Empirical Approach to Select Module Designs

Abstract

In modular robots, the shape of the building blocks (robotic modules) greatly influences the end result. By changing the physical properties of the module, different robotic structures with better performance for a given task can be found. In this paper, we modify the modules of a modular robot platform, the EMERGE modular robot, in two different ways: changing the length of the module and changing the shape of the starting module (base). We use artificial evolution to optimize robots for a locomotion task using each different module length and base, and also evolve robots with combinations of modules of different length. Results show that, as the length of the module increases, the best robots obtained use fewer modules and fewer connections per module. However, the increase in length results also in a decrease in locomotion performance for large length increases. Interestingly, very few of the best robots found show symmetric structures, which can be attributed to their tendency to roll over as their main means of locomotion. Modular robot designers can use the information about the effectiveness of modules with different lengths, and the use of different starting bases, to reach trade-offs between the desired number of modules in a robot and their effectiveness for a given task.