Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Retrospective on the Robot Air Hockey Challenge: Benchmarking Robust, Reliable, and Safe Learning Techniques for Real-world Robotics

Puze Liu, Jonas Günster, Niklas Funk, Simon Gröger, Dong Chen, Haitham Bou-Ammar, Julius Jankowski, Ante Marić, Sylvain Calinon, Andrej Orsula, Miguel Olivares-Mendez, Hongyi Zhou, Rudolf Lioutikov, Gerhard Neumann, Amarildo Likmeta Amirhossein Zhalehmehrabi, Thomas Bonenfant, Marcello Restelli, Davide Tateo, Ziyuan Liu, Jan Peters

TL;DR

The competition's results show that solutions combining learning-based approaches with prior knowledge outperform those relying solely on data when real-world deployment is challenging, and the ablation study reveals which real-world factors may be overlooked when building a learning-based solution.

Abstract

Machine learning methods have a groundbreaking impact in many application domains, but their application on real robotic platforms is still limited. Despite the many challenges associated with combining machine learning technology with robotics, robot learning remains one of the most promising directions for enhancing the capabilities of robots. When deploying learning-based approaches on real robots, extra effort is required to address the challenges posed by various real-world factors. To investigate the key factors influencing real-world deployment and to encourage original solutions from different researchers, we organized the Robot Air Hockey Challenge at the NeurIPS 2023 conference. We selected the air hockey task as a benchmark, encompassing low-level robotics problems and high-level tactics. Different from other machine learning-centric benchmarks, participants need to tackle practical challenges in robotics, such as the sim-to-real gap, low-level control issues, safety problems, real-time requirements, and the limited availability of real-world data. Furthermore, we focus on a dynamic environment, removing the typical assumption of quasi-static motions of other real-world benchmarks. The competition's results show that solutions combining learning-based approaches with prior knowledge outperform those relying solely on data when real-world deployment is challenging. Our ablation study reveals which real-world factors may be overlooked when building a learning-based solution. The successful real-world air hockey deployment of best-performing agents sets the foundation for future competitions and follow-up research directions.

A Retrospective on the Robot Air Hockey Challenge: Benchmarking Robust, Reliable, and Safe Learning Techniques for Real-world Robotics

TL;DR

The competition's results show that solutions combining learning-based approaches with prior knowledge outperform those relying solely on data when real-world deployment is challenging, and the ablation study reveals which real-world factors may be overlooked when building a learning-based solution.

Abstract

Machine learning methods have a groundbreaking impact in many application domains, but their application on real robotic platforms is still limited. Despite the many challenges associated with combining machine learning technology with robotics, robot learning remains one of the most promising directions for enhancing the capabilities of robots. When deploying learning-based approaches on real robots, extra effort is required to address the challenges posed by various real-world factors. To investigate the key factors influencing real-world deployment and to encourage original solutions from different researchers, we organized the Robot Air Hockey Challenge at the NeurIPS 2023 conference. We selected the air hockey task as a benchmark, encompassing low-level robotics problems and high-level tactics. Different from other machine learning-centric benchmarks, participants need to tackle practical challenges in robotics, such as the sim-to-real gap, low-level control issues, safety problems, real-time requirements, and the limited availability of real-world data. Furthermore, we focus on a dynamic environment, removing the typical assumption of quasi-static motions of other real-world benchmarks. The competition's results show that solutions combining learning-based approaches with prior knowledge outperform those relying solely on data when real-world deployment is challenging. Our ablation study reveals which real-world factors may be overlooked when building a learning-based solution. The successful real-world air hockey deployment of best-performing agents sets the foundation for future competitions and follow-up research directions.

Paper Structure

This paper contains 46 sections, 15 equations, 11 figures, 15 tables.

Table of Contents

  1. Introduction
  2. The Robot Air Hockey Challenge
  3. Competition Structure
  4. Qualifying stage
  5. Tournament stage
  6. Real-world validation stage
  7. Metrics
  8. Analysis
  9. Qualifying stage
  10. Ablation studies
  11. Tournament stage
  12. Real robot deployment
  13. Participants solution analysis
  14. Key research problems
  15. Limitations
  16. ...and 31 more sections

Figures (11)

  • Figure 1: Game played in the real world between SpaceR and Air-HocKIT. The Air-HocKIT robot (back) hits the puck and the SpaceR robot (front) defends the attack to take control of the puck.
  • Figure 2: The three tasks in the qualifying stage of the Robot Air Hockey Challenge: from left to right "Defend", "Hit", and "Prepare" tasks. The "Defend" task requires stopping an incoming puck to get control of it. The "Hit" task consists of scoring a goal against an opponent, which moves in a fixed pattern. The "Prepare" task consists of repositioning the puck in the central area of the table without losing control of it.
  • Figure 3: The tournament stage. The left figure shows the AiRLIHockey agent (right side of the table) performing a hitting motion, while the right figure shows the Air-HocKIT (left side of the table) defending the attack and taking control of the puck
  • Figure 4: Ablation studies on the effect of each domain discrepancy on the submitted solutions. The graph reports the success rate for each task under different types of environment modifications
  • Figure 5: Control diagram of the simulated environment
  • ...and 6 more figures