Imperfect Demonstration

Imperfect demonstration in imitation learning addresses the challenge of training robots or AI agents using suboptimal or noisy human demonstrations, a common scenario in real-world applications. Current research focuses on developing algorithms that can effectively filter or weight imperfect data, leveraging techniques like behavior cloning, adversarial training, and confidence-based methods to learn robust policies. These advancements are crucial for enabling robots to learn complex tasks from readily available, but imperfect, human guidance, thereby reducing the need for extensive, perfectly curated datasets. This work has significant implications for accelerating the development of autonomous systems across various domains.

Papers

June 13, 2024

A Dual Approach to Imitation Learning from Observations with Offline Datasets
Harshit Sikchi, Caleb Chuck, Amy Zhang, Scott Niekum
Imitation Learning Evolving Observation Offline Data Expert Level Performance Downstream Policy Learning Dual Approach Inverse Dynamic Imperfect Demonstration

May 24, 2024

How to Leverage Diverse Demonstrations in Offline Imitation Learning
Sheng Yue, Jiani Liu, Xingyuan Hua, Ju Ren, Sen Lin, Junshan Zhang, Yaoxue Zhang
Behavior Cloning Noisy Demonstration Expert Demonstration Offline Imitation Learning Imperfect Demonstration

January 17, 2024

SWBT: Similarity Weighted Behavior Transformer with the Imperfect Demonstration for Robotic Manipulation
Kun Wu, Ning Liu, Zhen Zhao, Di Qiu, Jinming Li, Zhengping Che, Zhiyuan Xu, Qinru Qiu, Jian Tang
Imitation Learning Robotic Manipulation Manipulation Task Expert Demonstration Offline Imitation Imperfect Demonstration Behavior Transformer

December 18, 2023

Aligning Human Intent from Imperfect Demonstrations with Confidence-based Inverse soft-Q Learning
Xizhou Bu, Wenjuan Li, Zhengxiong Liu, Zhiqiang Ma, Panfeng Huang
Fine Grained Imitation Learning Policy Learning Human Intent Soft Q Learning Imperfect Demonstration

October 30, 2023

Adversarial Batch Inverse Reinforcement Learning: Learn to Reward from Imperfect Demonstration for Interactive Recommendation
Jialin Liu, Xinyan Su, Zeyu He, Xiangyu Zhao, Jun Li
Reinforcement Learning Inverse Reinforcement Learning Personalized Recommendation Adversarial Inverse Reinforcement Learning Interactive Recommender System Imperfect Demonstration

October 11, 2023

Imitation Learning from Purified Demonstration
Yunke Wang, Minjing Dong, Bo Du, Chang Xu
Imitation Learning Imperfect Demonstration Optimal Demonstration

September 7, 2023

Improving Open Information Extraction with Large Language Models: A Study on Demonstration Uncertainty
Chen Ling, Xujiang Zhao, Xuchao Zhang, Yanchi Liu, Wei Cheng, Haoyu Wang, Zhengzhang Chen, Takao Osaki, Katsushi Matsuda, Haifeng Chen, Liang Zhao
Large Language Model Study Feature Unstructured Text Open Information Extraction Imperfect Demonstration

March 1, 2023

MEGA-DAgger: Imitation Learning with Multiple Imperfect Experts
Xiatao Sun, Shuo Yang, Mingyan Zhou, Kunpeng Liu, Rahul Mangharam
Imitation Learning Multi Expert Alignment Distillation Imperfect Demonstration

September 21, 2022

Goal-Aware Generative Adversarial Imitation Learning from Imperfect Demonstration for Robotic Cloth Manipulation
Yoshihisa Tsurumine, Takamitsu Matsubara
Generative Adversarial Imitation Learning Robotic Cloth Manipulation Imperfect Demonstration

February 7, 2022

Learning from Imperfect Demonstrations via Adversarial Confidence Transfer
Zhangjie Cao, Zihan Wang, Dorsa Sadigh
LeArning Abstract Expert Demonstration Prediction Confidence Algorithm Representation Imperfect Demonstration Adversarial Matching Method

Imperfect Demonstration

Papers

A Dual Approach to Imitation Learning from Observations with Offline Datasets

How to Leverage Diverse Demonstrations in Offline Imitation Learning

SWBT: Similarity Weighted Behavior Transformer with the Imperfect Demonstration for Robotic Manipulation

Aligning Human Intent from Imperfect Demonstrations with Confidence-based Inverse soft-Q Learning

Adversarial Batch Inverse Reinforcement Learning: Learn to Reward from Imperfect Demonstration for Interactive Recommendation

Imitation Learning from Purified Demonstration

Improving Open Information Extraction with Large Language Models: A Study on Demonstration Uncertainty

MEGA-DAgger: Imitation Learning with Multiple Imperfect Experts

Goal-Aware Generative Adversarial Imitation Learning from Imperfect Demonstration for Robotic Cloth Manipulation

Learning from Imperfect Demonstrations via Adversarial Confidence Transfer