Development and Integration of Self-Adaptation Strategies for Robotics Software

Robots are becoming more prevalent in industry and society as a whole. Alongside this growth their application domain is also broadening. Each application brings with it a host of potential uncertainties that the robots should be able to handle at runtime. To tackle this, the doctoral thesis outlined in this paper proposes to address three main problems. First, the current ad-hoc state of robotics software which impedes its evolution. Second, the inability to imagine every possible uncertainty at design time leading to unexpected scenarios at runtime. Third, unexpected scenarios resulting from the reality gap between the simulated environments in which robots are developed versus the real world. These unexpected scenarios may cause a system to violate its requirements, especially in our case non-functional requirements. In an attempt to solve these problems, we plan to implement a variety of self-adaptation strategies. These strategies allow systems to change their composition to handle the afore-mentioned unexpected events during operation autonomously. To accomplish this we will need to reason about how best to integrate these strategies into the software of existing robots, as well as how existing information available to designers regarding the robots can best be utilized to improve the strategies. Lastly, these strategies and the process through which they are integrated will be assessed in their impact across different robotic case studies. Preliminary results from the work towards the thesis are also presented, alongside a consideration of its potential industrial impact.