Intention Recognition and Object Recommendation System using Deep Auto-encoder Based Affordance Model

Intention recognition is an important task for human-agent interactions (HAI) since it can make the robot respond adequately to the human’s intention. For the robot to understand the world in terms of its own actions, the robot requires the definition of adequate knowledge representations. Affordance is the concept used to represent the relation between an agent and its environment. A robot can exploit this type of knowledge to infer implicit human intentions. In this paper, we propose a system based on action-object affordances modeled using deep structure that can recognize the user’s intention and recommend the corresponding objects related to that intention. The network is learnt by the robot after considering the user’s attention for specific objects. To notice the user’s attention, the gaze information is obtained using Tobii 1750 eye-tracker in experiments. The experimental results show the successful recognition and recommendation performance of the proposed system.