PPHR: A Personalized AI System for Proactive Robots

This paper proposes the Personalized Proactive Home Robotic (PPHR) system: an intelligent robotics system that enables proactive behavior in the context of a home environment. With the implementation of the PPHR system, robots will be able to predict actions that users would want by gathering data about both the users and the environment. Then, once confident predictions are made, the robot can perform the actions either actively or proactively depending on the situation. The system also uses personalized learning models to adapt the experience to each of its users, and federated learning to improve data privacy and train models faster. To prove the viability of the system, we have designed and implemented the learning capabilities using data gathered internally. Additionally, we have shown the transfer learning capabilities of the system, allowing users to actively add new actions to the robot at any time. With promising results, the system will serve as a large step towards improved human-robot interaction.