SoPhie: An Attentive GAN for Predicting Paths Compliant to Social and Physical Constraints
share
We are witnessing an increasing growth of autonomous platforms such as self-driving cars and social robots around us. These mobile machines need to navigate dynamic human environments safely and in a socially acceptable way. Hence, understanding the human behavior and the physical constraints of the environment is crucial. To address the task of path prediction for multiple interacting agents in a scene, we present SoPhie; an interpretable framework based on Generative Adversarial Network (GAN). Sophie leverages two sources of information: the past trajectory of all the agents in the scene and a wide top-view image of the navigation scene. Our approach combines a physical attention mechanism with a social attention mechanism. The physical attention component helps the model to learn where to look in a large scene and extract the most salient parts of the image relevant to the path. Whereas, the social attention component is able to aggregate information across different involved agents and extract the most important trajectory information from the surrounding agents, enabling our method to predict socially aware paths. SoPhie is able to capture the multi-modal nature of the future prediction: given the history of trajectories and a navigation scene, there can be multiple acceptable paths in the future. Through experiments on several trajectory forecasting benchmarks, we show that SoPhie outperforms prior work in terms of accuracy, while predicting physically appropriate paths. We also show that SoPhie is able to classify traversable vs non-traversable locations in a scene.
READ FULL TEXT
