Publications

Fitting Reinforcement Learning Model to Behavioral Data under Bandits

Investigates fitting RL models to behavioral data under bandit feedback, proposing a new method and demonstrating its effectiveness.

Constrained Reinforcement Learning with Smoothed Log Barrier Function

Proposes a Smoothed Log Barrier Function (CSAC-LB) for constrained RL, proving convergence to the optimal policy and demonstrating effectiveness on safety-critical tasks.

Goal Achievement Guided Exploration: Mitigating Premature Convergence in Learning Robot Control

Proposes Goal Achievement Guided Exploration (GAGE) to mitigate premature convergence and improve sample efficiency in learning robot control with sparse rewards.

Revisiting Safe Exploration in Safe Reinforcement learning

Introduces a new safety metric, expected maximum consecutive cost steps (EMCC), to address limitations in standard SafeRL cost metrics and improve safe exploration.

Constrained Reinforcement Learning for Safe Heat Pump Control

Applies Constrained Reinforcement Learning to heat pump control, minimizing energy consumption while ensuring comfort and safety.

Learning Continuous Control with Geometric Regularity from Robot Intrinsic Symmetry

Introduces novel network structures for single-agent robot control that explicitly capture inherent reflectional and rotational symmetries to improve learning.

Automated Reinforcement Learning (AutoRL): A Survey and Open Problems

A comprehensive survey of Automated Reinforcement Learning (AutoRL), unifying the field and discussing its challenges and potential.

Automated reinforcement learning (autorl): A survey and open problems

A comprehensive survey of Automated Reinforcement Learning (AutoRL), unifying the field and discussing its challenges and potential.

On the Importance of Hyperparameter Optimization for Model-based Reinforcement Learning

Investigating the impact of Hyperparameter Optimization on Model-based RL performance.