• Classified by Topic • Classified by Publication Type • Sorted by Date • Sorted by First Author Last Name • Classified by Funding Source •
Deep Reinforcement Learning for Robotics: A Survey of Real-World Successes.
Chen Tang, Ben Abbatematteo, Jiaheng
Hu, Rohan Chandra, Roberto Martín-Martín, and Peter Stone.
Annual
Review of Control, Robotics, and Autonomous Systems (ARCRAS), 2024.
[PDF]4.3MB [slides.pdf]2.9MB [poster.pdf]604.3kB
Reinforcement learning (RL), particularly its combination with deep neuralnetworks referred to as deep RL (DRL), has shown tremendous promise across a widerange of applications, suggesting its potential for enabling the development ofsophisticated robotic behaviors. Robotics problems, however, pose fundamentaldifficulties for the application of RL, stemming from the complexity and cost ofinteracting with the physical world. This article provides a modern survey of DRLfor robotics, with a particular focus on evaluating the real-world successesachieved with DRL in realizing several key robotic competencies. Our analysisaims to identify the key factors underlying those exciting successes, revealunderexplored areas, and provide an overall characterization of the status of DRLin robotics. We highlight several important avenues for future work, emphasizingthe need for stable and sample-efficient real-world RL paradigms, holisticapproaches for discovering and integrating various competencies to tackle complexlong-horizon, open-world tasks, and principled development and evaluationprocedures. This survey is designed to offer insights for both RL practitionersand roboticists toward harnessing RL's power to create generally capablereal-world robotic systems.
@Article{chen_tang_ARCRAS2024, author = {Chen Tang and Ben Abbatematteo and Jiaheng Hu and Rohan Chandra and Roberto MartÃn-MartÃn and Peter Stone}, title = {Deep Reinforcement Learning for Robotics: A Survey of Real-World Successes}, journal = {Annual Review of Control, Robotics, and Autonomous Systems (ARCRAS)}, year = {2024}, abstract = {Reinforcement learning (RL), particularly its combination with deep neural networks referred to as deep RL (DRL), has shown tremendous promise across a wide range of applications, suggesting its potential for enabling the development of sophisticated robotic behaviors. Robotics problems, however, pose fundamental difficulties for the application of RL, stemming from the complexity and cost of interacting with the physical world. This article provides a modern survey of DRL for robotics, with a particular focus on evaluating the real-world successes achieved with DRL in realizing several key robotic competencies. Our analysis aims to identify the key factors underlying those exciting successes, reveal underexplored areas, and provide an overall characterization of the status of DRL in robotics. We highlight several important avenues for future work, emphasizing the need for stable and sample-efficient real-world RL paradigms, holistic approaches for discovering and integrating various competencies to tackle complex long-horizon, open-world tasks, and principled development and evaluation procedures. This survey is designed to offer insights for both RL practitioners and roboticists toward harnessing RL's power to create generally capable real-world robotic systems. }, }
Generated by bib2html.pl (written by Patrick Riley ) on Thu Mar 20, 2025 13:36:49