Context-Aware Proactive Robotic Companions for Care Homes and Communities

The question of whether robots have a place in healthcare is one that is becoming more and more necessary to consider. Shortages of healthcare staff, lack of funding for healthcare services and the continuously growing population have begun to highlight the increasing need for changes in the way we help vulnerable members of our society. In this thesis, I have explored the potential of robots, specifically the humanoid robot Pepper, to help fill this gap when it comes to the care of the elderly who may have a diverse range of needs by providing users with aid and companionship in their daily lives whether that is retaining independence for longer at home, or as part of a larger care community. I have verified that through the development and implementation of a biologically inspired cognitive architecture onto the Pepper robot, it is possible to create a robotic companion that can be of assistance to patients and carers alike, by completing requests and engaging in social interaction with users and in turn, reducing the workload of carers. The following chapters document this through a series of practical and simulated experiments that have formed multiple software modules including a cumulatively growing artificial episodic memory system, a human-aware navigation system and a social intelligence module. Through this, Pepper has been able to successfully learn and recall experiences while accounting for current social and environmental contexts and interact with users through a multimodal interface to provide a sense of independence and personalised companionship. The architecture also includes a swift-learning human-aware navigation system capable of navigating and mapping unstructured environments. This method has enabled Pepper to quickly map/re-map new and previously visited environments with average training of single rooms taking up to 120 seconds on average and larger rooms (> 10,000 data points) taking up to 350 seconds.