Building A Socially Acceptable Navigation and Behavior of A Mobile Robot Using Q-Learning

This paper proposes a self-learner mobile robot to navigate among social environment under social force framework. It addresses a drawback of the Social Force Navigation Model (SFNM) based approaches which used a set of fixed-SFNM parameters that is not always in accordance with various conditions. Those fixed parameters are usually assumed as the most optimal generated parameters by optimizing a collection of data samples from the specific interaction of human beings. We utilize the Qlearning algorithm to select adaptively the parameters of SFNM to deal with each circumstance. It implies that we train our robot to interact with the environments directly. However, training the real robot in the real environments under Q-learning framework is difficult, time-consuming, and hazardous. Therefore, in this study, we utilize a realistic simulator, V-Rep, for both training and testing. The simulation results for several scenarios exhibit the usefulness of our approach to smoothly and safely navigate our robot from start to the goal position.