A Framework for Service Robots in Smart Home: An Efficient Solution for Domestic Healthcare

Abstract Purpose As the elder population grows, the need for domestic healthcare is on the rise. Both robotics and smart environments, including smart homes, provide a promising solution to monitor, interact and keep company to users. However, in real case scenarios, sensors data are not perfect and the environment changes over time, leading to erroneous understanding of the context and inappropriate responses. The purpose of this work is to tackle those challenges in order to improve the autonomy and efficiency of robots in smart environments. Methods The problematic was structured into three steps: (1) perception, (2) cognition and (3) action. We proposed and evaluated a software framework that covers the challenges of each step. It includes respectively: (1) a context acquisition method that supports and models the uncertainty of data by using complex event processing, fuzzy logic and ontologies; (2) an activity recognition system that combines vision, context knowledge and semantic reasoning; (3) a dynamic hierarchical task planner that alternates planning and execution. For each step, the framework was evaluated through simulations and/or experiments using a robot and a smart room. Results The quality of the perception was assessed by measuring the efficiency of a cognition process using the acquired context knowledge. An uncertain environment was simulated, and results show our framework to enable a gain of 10% of correctness for an activity recognition process. The cognition part of the framework was evaluated by observing several persons performing activities. It achieved an overall 90% correct recognition, yet, such result questions the relevance of our approach. Finally, the action step was confronted a simulated scenario with various levels of dynamism. Our task planner appeared to reduce, by up to 23%, the number of tasks required to reach a goal in a dynamic environment. Conclusion Our framework provides software tools that make robots and smart environments more relevant in real housings. By supporting the uncertainty of context data and the dynamism of the environment, robots and smart environments can achieve more effectively their purposes in domestic healthcare applications.