Hydra-MDP: End-to-end Multimodal Planning with Multi-target Hydra-Distillation

TL;DR

Hydra-MDP introduces a multi-teacher, end-to-end multimodal planning framework that distills knowledge from human and rule-based teachers, achieving state-of-the-art results in autonomous driving challenges.

Contribution

It presents a novel multi-teacher distillation approach with a multi-head decoder for diverse trajectory generation in autonomous driving.

Findings

Achieved 1st place in Navsim challenge.

Demonstrated improved generalization across environments.

Learned environment influence end-to-end without non-differentiable post-processing.

Abstract

We propose Hydra-MDP, a novel paradigm employing multiple teachers in a teacher-student model. This approach uses knowledge distillation from both human and rule-based teachers to train the student model, which features a multi-head decoder to learn diverse trajectory candidates tailored to various evaluation metrics. With the knowledge of rule-based teachers, Hydra-MDP learns how the environment influences the planning in an end-to-end manner instead of resorting to non-differentiable post-processing. This method achieves the $1^{st}$ place in the Navsim challenge, demonstrating significant improvements in generalization across diverse driving environments and conditions. More details by visiting \url{https://github.com/NVlabs/Hydra-MDP}.