This video is part of the paper: "Efficient and Robust LiDARBased EndtoEnd Navigation" which is presented at the International Conference on Robotics and Automation (ICRA) 2021.
Efficient and Robust LiDARBased EndtoEnd Navigation
Zhijian Liu^, Alexander Amini^, Sibo Zhu, Sertac Karaman, Song Han, and Daniela L. Rus
Paper Link: https://arxiv.org/abs/2105.09932
Website: https://le2ed.mit.edu
Abstract:
Deep learning has been used to demonstrate endtoend neural network learning for autonomous vehicle control from raw sensory input. While LiDAR sensors provide reliably accurate information, existing endtoend driving solutions are mainly based on cameras since processing 3D data requires a large memory footprint and computation cost. On the other hand, increasing the robustness of these systems is also critical; however, even estimating the model’s uncertainty is very challenging due to the cost of samplingbased methods. In this paper, we present an efficient and robust LiDARbased endtoend navigation framework. We first introduce FastLiDARNet that is based on sparse convolution kernel optimization and hardwareaware model design. We then propose Hybrid Evidential Fusion that directly estimates the uncertainty of the prediction from only a single forward pass and then fuses the control predictions intelligently. We evaluate our system on a fullscale vehicle and demonstrate lanestable as well as navigation capabilities. In the presence of outofdistribution events (e.g., sensor failures), our system significantly improves robustness and reduces the number of takeovers in the real world.