Motion planning for modular robots under failures

In this thesis, methods of motion planning for modular robots under failure is investigated. Methods based on Central Pattern Generator (CPG) are used to generate signals for locomotion control of modules. The Particle Swarm Optimization (PSO) is bio-inspired method which is used for optimization of CPG-based locomotion. For purpose of the motion planning, the Rapidly-exploring Random Tree (RRT) algorithm is implemented. In this thesis, two types of module failure are investigated, fix-joint and free-joint. The influence of different combinations of modules under failures was statistically compared on different types of robots. The experiments were conducted using a simulation environment. The result of the experiments is that in most of the cases robots operate better with broken modules as without them.