Learning and Development of Home Service Robots' Service Cognition Based on a Learning Mechanism

In order to improve the intelligence of home service robots and resolve their inability to develop service cognition skills in an autonomous, human-like manner, we propose a method for home service robots to learn and develop skills that allow them to perform their services appropriately in a dynamic and uncertain home environment. In a context model built with the support of intelligent sensors and Internet of Things (IoT) technology in a smart home, common-sense information about environmental comfort is recorded into the logical judgment of the robot as a reward provided by the environment. Our approach uses a reinforcement learning algorithm that helps train the robot to provide appropriate services that bring the environment to the user's comfort level. We modified the incremental hierarchical discriminant regression (IHDR) algorithm to construct an IHDR tree from the discrete part of the data in a smart home to store the robot's historical experience for further service cognition. Poor adaptive capacity in a changeable home environment is avoided by additional user guidance, which can be inputted after the decision is made by the IHDR tree. In the early development period, when robots make an inappropriate service decision because they lack historical experience, the user can help fix this decision. Then, the IHDR tree is updated incrementally with fixed decisions to enrich the robot's empirical knowledge and realize the development of its autonomic cognitive ability. The experimental results show that the robot accumulates increasingly more experience over time, and this experience plays an important role in its future service cognition, similar to the process of human mental development.