Justin Lidard

Welcome! I’m a PhD candidate at Princeton, where I work with Anirudha Majumdar and Naomi Leonard on robotics. During my PhD I’ve spent time at Toyota Research Institute working on human-interactive driving.
I design algorithms that enable robots to intelligently interact with and learn from people. One of my key interests is value alignment from an optimal control perspective: AI systems may offer strong performance but lack nuance in their understanding of intuitive and safe behaviors. To bridge this gap, I am working to enable robots to closely mimic people through imitation while optimizing their long-term behavior for safety and robustness. My long-term goal is to enable robots to recognize their own uncertainty and continuously adjust their strategies through feedback during training, evaluation, and deployment.
jlidard
at princeton
dot edu
News
Feb 15, 2025 | Really exciting new work on scaling up safety filters to high-dimensional, human-in-the-loop systems coming soon. Stay tuned! |
---|---|
Feb 01, 2025 | DPPO is accepted to ICLR 2025. See you in Singapore! |
Dec 01, 2024 | New survey paper on uncertainty quantification in LLMs. |
Sep 05, 2024 | New preprint with Allen Ren on fine-tuning Diffusion Policies with PPO! |
Mar 05, 2024 | Two papers (KLGame, RCIP) accepted to RSS 2024. See you in Delft! |
Selected Publications
- Safety with Agency: Human-Centered Safety Filter with Application to AI-Assisted MotorsportsUnder Review., 2025
-
-
- Risk-Calibrated Human-Robot Interaction via Set-Valued Intent PredictionRobotics: Science and Systems (RSS), 2024