Evaluating Autonomous Systems Before Deployment

← Back to Research Overview

Autonomous systems must operate safely in complex human environments, but real-world testing is limited and risky. We develop scalable evaluation methods — including generative simulation for interactive environments, diffusion models for trajectory prediction and scenario generation, and off-policy evaluation techniques — to rigorously assess autonomous system performance before deployment. We also build benchmarks and datasets for driving reasoning to enable standardized evaluation across the community.

Selected Publications

2025

  1. ICML 2025
    womd.png
    WOMD-Reasoning: A Large-Scale Language Dataset for Interactions and Driving Intentions Reasoning
    Yiheng Li, Cunxin Fan, Chongjian Ge, and 9 more authors
    Forty-Second International Conference on Machine Learning (ICML), 2025

2024

  1. ECCV 2024
    ogd.png
    Optimizing Diffusion Models for Joint Trajectory Prediction and Controllable Generation
    Yixiao Wang, Chen Tang, Lingfeng Sun, and 8 more authors
    In European Conference on Computer Vision (ECCV), 2024

2021

  1. NeurIPS 2021
    social.png
    Exploring Social Posterior Collapse in Variational Autoencoder for Interaction Modeling
    Chen Tang, Wei Zhan, and Masayoshi Tomizuka
    In Advances in Neural Information Processing Systems (NeurIPS), 2021