Zizhao Wang

I am a PhD student in ECE at the University of Texas at Austin, advised by Prof. Peter Stone. My research focuses on the intersection of causality and decision-making (mostly reinforcement learning).

Previously, I completed my M.S. in CS at Columbia University, advised by Prof. Peter Allen and Prof. Itsik Pe’er, and my undergraduate studies at the University of Michigan - Ann Arbor.

Email  /  Scholar  /  Twitter  /  CV

profile photo

Selected Works

SkiLD: Unsupervised Skill Discovery Guided by Factor Interactions
Zizhao Wang*, Jiaheng Hu*, Caleb Chuck*, Stephen Chen, Roberto Martín-Martín, Amy Zhang, Scott Niekum, Peter Stone
preprint, 2024
paper / code / project page

A unsupervised skill discovery method that induces diverse interactions between state factors, which are often more valuable for solving downstream tasks.

CaMP: Causal Motion Predictor for Robust Trajectory Forecasting
Zizhao Wang*, Chen Tang, Aolin Xu, Enna Sachdeva, Peter Stone, Teruhisa Misu
preprint, 2024
paper (to be released)

Improve the robustness of motion prediction models by inferring causal relationships between agents.

Disentangled Unsupervised Skill Discovery for Efficient Hierarchical Reinforcement Learning
Jiaheng Hu, Zizhao Wang, Peter Stone, Roberto Martín-Martín
preprint, 2024
paper / code / project page

A method for learning disentangled skills where each skill component only affects one factor of the state space, so skills can be efficiently reused to solve downstream tasks.

Building Minimal and Reusable Causal State Abstractions for Reinforcement Learning
Zizhao Wang*, Caroline Wang*, Xuesu Xiao, Yuke Zhu, Peter Stone,
AAAI, 2024 (Oral)
paper / slides

Improve exploration by visiting states where the agent is uncertain whether (as opposed to how) entities such as the agent or objects have some influence on each other.

ELDEN: Exploration via Local Dependencies
Jiaheng Hu*, Zizhao Wang*, Peter Stone, Roberto Martín-Martín
Neurips, 2023
paper / poster

Learn the causal relationships in the dynamics and reward functions for each task to derive a minimal, task-specific abstraction.

Learning to Correct Mistakes: Backjumping in Long-Horizon Task and Motion Planning
Yoonchang Sung*, Zizhao Wang*, Peter Stone
CoRL, 2022
paper

When task and motion planning search reaches dead-ends due to an early action, we identify and backjump to the culprit action to resample it, improving the search efficiency compared to backtracking.

Causal Dynamics Learning for Task-Independent State Abstraction
Zizhao Wang, Xuesu Xiao, Zifan Xu, Yuke Zhu, Peter Stone,
ICML, 2022 (Oral)
paper / poster / slides / code

Learn a theoretically proved causal dynamics model that removes unnecessary dependencies between state variables and the action, so it generalizes well to unseen states.

From agile ground to aerial navigation: Learning from learned hallucination
Zizhao Wang, Xuesu Xiao, Alexander J Nettekoven, Kadhiravan Umasankar, Anika Singh, Sriram Bommakanti, Ufuk Topcu, Peter Stone,
IROS, 2021
paper / poster / slides

Generate cheap training data for navigation by hallucinating obstacles.

APPLE: Adaptive Planner Parameter Learning from Evaluative Feedback
Zizhao Wang, Xuesu Xiao, Garrett Warnell, Peter Stone,
IROS, 2021
paper / slides

Learn how to dynamically adjust planner parameters using evaluative feedback from non-expert users.

APPLI: Adaptive Planner Parameter Learning from Interventions
Zizhao Wang, Xuesu Xiao, Bo Liu, Garrett Warnell, Peter Stone,
ICRA, 2021
paper / slides

Learn how to dynamically adjust planner parameters using interventions from non-expert users.

Variational Objectives for Markovian Dynamics with Backward Simulation
Antonio Khalil Moretti*, Zizhao Wang*, Luhuan Wu∗, Iddo Drori, Itsik Pe’er,
ECAI, 2020
paper

A novel variational inference framework for nonlinear hidden Markov models.

Teaching and Services

Organizer of Causality for Robotics Workshop at IROS 2023
Reviewer for NeurIPS, ICML, IROS, ICRA, and RA-L
Teaching assistant for Reinforcement Learning (CS 394R), Causality and Reinforcement Learning (ECE 381V), and Artificial Intelligence (CS 343).