Show HN: Rally (Real-Time Dual-Arm Robotic Table Tennis System) (github.com)

Rally, an innovative dual-arm robotic table tennis system, leverages high-performance real-time control with C++ and Python orchestration to simulate engaging matches. The system operates at a rapid frequency of 1kHz for physics dynamics and 500Hz per arm, showcasing advanced technologies such as deterministic ownership arbitration, an EKF-based ball prediction algorithm, and recursive least squares for adaptive strategy. Key components—including a physics arbiter for ball dynamics and ownership control, and an analytical inverse kinematics solution for arm movements—highlight Rally's robust architecture, achieving a notable ball prediction error of just 4.1 mm when validated against real robot data. This development is significant for the AI/ML community as it exemplifies the integration of machine learning with real-time robotics, presenting a practical application through a competitive simulation. The architecture utilizes ZeroMQ for seamless communication, ensuring both process isolation and determinism vital for real-time performance. Additionally, Rally's potential for future integration with large language models (LLMs) for tactical analysis presents exciting implications for enhancing human-robot interaction in sports settings. Overall, Rally not only pushes the boundaries of robotic control systems but also introduces a novel platform for testing and refining algorithms in dynamic environments.