Robotic systems that are able to adapt to individual humans through long-term or life-long experience and learning pave the way towards closer collaboration and interaction in a variety of application domains. In my talk I will outline recent scientific advances towards this aim, accomplished in a number of different robotics projects. I will discuss the potential benefits of long-term autonomy and adaptation as well as its challenges and how they can be addressed in terms of machine learning, software engineering, human-robot interaction, and artificial intelligence. Application domains to be presented in my talk will range from service applications of mobile robots in care and security, to individualised collaboration between manipulating robots and humans in manufacturing settings.
In particular, I'll show examples from our research covering the perception, interpretation, and adaptation of human motion and activity (using qualitative spatial relations), probabilistic models of task engagement and prediction, as well as continuous integration and testing in robotic software development.