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

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.

Energy-based Contact Planning under Uncertainty for Robot Air Hockey

Planning robot contact often requires reasoning over a horizon to anticipate outcomes, making such planning problems computationally expensive. In this letter, we propose a learning framework for efficient contact planning in real-time subject to uncertain contact dynamics. We implement our approach for the example task of robot air hockey. Based on a learned stochastic model of puck dynamics, we formulate contact planning for shooting actions as a stochastic optimal control problem with a chance constraint on hitting the goal. To achieve online re-planning capabilities, we propose to train an energy-based model to generate optimal shooting plans in real time. The performance of the trained policy is validated %in experiments both in simulation and on a real-robot setup. Furthermore, our approach was tested in a competitive setting as part of the NeurIPS 2023 Robot Air Hockey Challenge.

AiRLIHockey: Highly Reactive Contact Control and Stochastic Optimal Shooting

Air hockey is a highly reactive game which requires the player to quickly reason over stochastic puck and contact dynamics. We implement a hierarchical framework which combines stochastic optimal control for planning shooting angles and sampling-based model-predictive control for continuously generating constrained mallet trajectories. Our agent was deployed and evaluated in simulation and on a physical setup as part of the Robot Air-Hockey challenge competition at NeurIPS 2023.

Online Learning of Piecewise Polynomial Signed Distance Fields for Manipulation Tasks

Reasoning about distance is indispensable for establishing or avoiding contact in manipulation tasks. To this end, we present an online method for learning implicit representations of signed distance using piecewise polynomial basis functions. Starting from an arbitrary prior shape, our approach incrementally constructs a continuous representation from incoming point cloud data. It offers fast access to distance and analytical gradients without the need to store training data. We assess the accuracy of our model on a diverse set of household objects and compare it to neural network and Gaussian process counterparts. Distance reconstruction and real-time updates are further evaluated in a physical experiment by simultaneously collecting sparse point cloud data and using the evolving model to control a manipulator.