EWareNet: Emotion Aware Human Intent Prediction and Adaptive Spatial Profile Fusion for Social Robot Navigation

We present EWareNet, a novel intent and affect-aware social robot navigation\nalgorithm among pedestrians. Our approach predicts the trajectory-based\npedestrian intent from gait sequence, which is then used for intent-guided\nnavigation taking into account social and proxemic constraints. We propose a\ntransformer-based model that works on commodity RGB-D cameras mounted onto a\nmoving robot. Our intent prediction routine is integrated into a mapless\nnavigation scheme and makes no assumptions about the environment of pedestrian\nmotion. Our navigation scheme consists of a novel obstacle profile\nrepresentation methodology that is dynamically adjusted based on the pedestrian\npose, intent, and affect. The navigation scheme is based on a reinforcement\nlearning algorithm that takes pedestrian intent and robot's impact on\npedestrian intent into consideration, in addition to the environmental\nconfiguration. We outperform current state-of-art algorithms for intent\nprediction from 3D gaits.\n