On the Road to Portability: Compressing End-to-End Motion Planner for Autonomous Driving

arXiv:2403.01238v1 Announce Type: new
Abstract: End-to-end motion planning models equipped with deep neural networks have shown great potential for enabling full autonomous driving. However, the oversized neural networks render them impractical for deployment on resource-constrained systems, which unavoidably requires more computational time and resources during reference.To handle this, knowledge distillation offers a promising approach that compresses models by enabling a smaller student model to learn from a larger teacher model. Nevertheless, how to apply knowledge distillation to compress motion planners …

abstract arxiv autonomous autonomous driving computational cs.cv deployment driving enabling motion planning networks neural networks planning portability reference resources systems them type