Abstract: In this paper, we propose an end-to-end self-driving network featuring a sparse attention module that learns to automatically attend to important regions of the input. The attention module specifically targets motion planning, whereas prior literature only applied attention in perception tasks. Learning an attention mask directly targeted for motion planning significantly improves the planner safety by performing more focused computation. Furthermore, visualizing the attention improves interpretability of end-to-end self-driving.
Authors: Bob Wei, Mengye Ren, Wenyuan Zeng, Ming Liang, Bin Yang, Raquel Urtasun (Uber ATG)