| Summary: | The number of people affected by natural disasters and the financial costs of recovery are increasing. To improve the survival rate, reduce the financial costs and reduce the risks posed to rescue workers, several authors have proposed using robotic modules to aid in the search and rescue and cleanup operations that follow natural disasters. Due to the potentially high-levels of adaptivity that they may provide, there is one class of system in particular which is often highlighted: self-reconfigurable modular robots. Self-reconfigurable modular robots are robotic modules which can physically connect with one another and dynamically alter their structural configuration through a process known as morphogenesis. For self-reconfigurable modular robotic systems to be effective in tasks like search and rescue they will need to be capable of surviving autonomously for long periods of time, whilst demonstrating high levels of reliability and fault tolerance. In this thesis, a novel approach is used to study the reliability of an existing morphogenesis algorithm, using techniques from the field of reliability engineering. The techniques are found to be effective in identifying problems with the algorithm and the findings are used to help develop new strategies for detecting faults and recovering from failures. Results from simulated and real robot experiments show that systems employing these strategies are able to recover significantly quicker and survive longer than systems which do not. The design of a new platform extension and algorithms for controlling the collective locomotion of robots equipped with the extension are also presented. Through the novel use of `virtual sensors', the same locomotion strategy is used to demonstrate implicit forms of self-assembly and self-reconfiguration. It is concluded that the long-term autonomy of self-reconfigurable modular robotic systems can be improved through the study and development of new and existing approaches to fault tolerant morphogenesis.
|
|---|
