Robots with internal models: A route to self-aware and hence safer robots

The aim of this chapter is to set out the case for building robots with internal models as a possible route toward achieving a level of functional self-awareness that would usefully extend the capabilities of autonomous robots. The chapter argues that these capabilities will lead to enhanced safety – especially in physical human robot interaction (pHRI) – and, perhaps also, toward ethical behaviour in autonomous robots. Indeed, the chapter will advance the argument that safe and ethical autonomous robots may not be achievable at all without mechanisms for self-awareness. Importantly, the ideas and mechanisms proposed in this chapter are intended to be realisable with current and near-future technology, i.e. using conventional computing platforms embedded within existing or buildable robot bodies, with existing devices for sensing and actuation. Thus, this chapter is primarily about how we might engineer practical self-awareness, for safer (and possibly ethical) robots in the near-term. This chapter will be less concerned with philosophical questions such as whether, or not, such robots are really self-aware, although we will touch upon the question of what behaviour might, if exhibited, be argued as evidence for 'as if' self- awareness.