Deep Inverse Reinforcement Learning

Deep Inverse Reinforcement Learning (Deep IRL) aims to infer the reward function that governs an agent's behavior by observing its actions, typically using demonstrations from an expert. Current research focuses on applying Deep IRL to diverse problems, including robot navigation (in both social and unstructured environments), autonomous driving, and route optimization, often employing Maximum Entropy Deep IRL (MEDIRL) and related algorithms. These methods leverage deep neural networks to model complex reward functions and improve the efficiency and generalizability of learned policies compared to traditional approaches. The resulting advancements have significant implications for various fields, enabling the development of more robust and human-like autonomous systems.

Papers

September 5, 2024

Discovering Cyclists' Visual Preferences Through Shared Bike Trajectories and Street View Images Using Inverse Reinforcement Learning
Kezhou Ren, Meihan Jin, Huiming Liu, Yongxi Gong, Yu Liu
Multi Source Roadside Perception Route Choice Deep Inverse Reinforcement Learning

June 15, 2023

Joint Path planning and Power Allocation of a Cellular-Connected UAV using Apprenticeship Learning via Deep Inverse Reinforcement Learning
Alireza Shamsoshoara, Fatemeh Lotfi, Sajad Mousavi, Fatemeh Afghah, Ismail Guvenc
Unmanned Aerial Vehicle Q Learning Inverse Reinforcement Learning Power Allocation Apprenticeship Learning Deep Inverse Reinforcement Learning Joint Motion Planning

September 16, 2022

SoLo T-DIRL: Socially-Aware Dynamic Local Planner based on Trajectory-Ranked Deep Inverse Reinforcement Learning
Yifan Xu, Theodor Chakhachiro, Tribhi Kathuria, Maani Ghaffari
Inverse Reinforcement Learning Social Navigation Multimodal Planning Deep Inverse Reinforcement Learning

July 7, 2022

Energy-based Legged Robots Terrain Traversability Modeling via Deep Inverse Reinforcement Learning
Lu Gan, Jessy W. Grizzle, Ryan M. Eustice, Maani Ghaffari
Inverse Reinforcement Learning Terrain Traversability Exteroceptive Perception Deep Inverse Reinforcement Learning

June 18, 2022

A deep inverse reinforcement learning approach to route choice modeling with context-dependent rewards
Zhan Zhao, Yuebing Liang
Value Laden Choice Choice Model Discrete Choice Deep Inverse Reinforcement Learning

January 18, 2022

Inducing Structure in Reward Learning by Learning Features
Andreea Bobu, Marius Wiggert, Claire Tomlin, Anca D. Dragan
Reward Report Reward Learning Structured Document Adaptive Behavior Deep Inverse Reinforcement Learning

January 17, 2022

Spatiotemporal Costmap Inference for MPC via Deep Inverse Reinforcement Learning
Keuntaek Lee, David Isele, Evangelos A. Theodorou, Sangjae Bae
Autonomous Driving Reward Function Inverse Reinforcement Learning Cost Map Deep Inverse Reinforcement Learning

Deep Inverse Reinforcement Learning

Papers

Discovering Cyclists' Visual Preferences Through Shared Bike Trajectories and Street View Images Using Inverse Reinforcement Learning

Joint Path planning and Power Allocation of a Cellular-Connected UAV using Apprenticeship Learning via Deep Inverse Reinforcement Learning

SoLo T-DIRL: Socially-Aware Dynamic Local Planner based on Trajectory-Ranked Deep Inverse Reinforcement Learning

Energy-based Legged Robots Terrain Traversability Modeling via Deep Inverse Reinforcement Learning

A deep inverse reinforcement learning approach to route choice modeling with context-dependent rewards

Inducing Structure in Reward Learning by Learning Features

Spatiotemporal Costmap Inference for MPC via Deep Inverse Reinforcement Learning