A Q-Learning based Method for Energy-Efficient Monitoring of Activities of Daily Living using a Wearable Device and a Companion Robot

In this paper, we proposed a Q-learning based algorithm to trade off the activity recognition accuracy, energy cost, privacy concerns and robot resource consumption in monitoring activities of daily living (ADLs) for elderly care using a wearable device and a companion robot. The robot uses a Q-learning algorithm to decide when to turn on what sensors for data collection by considering the above four criteria. The wearable device follows the robot's decision to turn on the relevant sensors to collect data which are sent back to the robot for activity recognition. First, we described the overall design of the system in which a wearable device and a companion robot collaborate for activity monitoring. Second we developed the Q-learning framework based on the knowledge of transition motions. Third, we trained the model and evaluated the proposed method using offline data and real time data collected from our smart home testbed. The results showed that the proposed method could recognize ADLs with high accuracy while saving about half energy compared with periodic methods. The results from real time evaluation showed that the activity detection rate is about 85%.