RaysTwins

RaysTwins: A Simulation and Training Platform for Next-Generation Robotics and Embodied AI RaysTwins is a comprehensive simulation and training RaysTwins is a comprehensive simulation and training platform designed for next-generation robotics and embodied AI. It integrates rapid scene modelling asset management interaction orchestration high-performance physics and sensor simulation large-scale parallel training. It integrates rapid scene modelling asset management interaction orchestration high-level fidelity physics and sensor simulation large-scale parallel training and native ROS 2 interoperability-enabling an end-to-end It integrates rapid scene modelling, asset management, interaction orchestration, high-level fidelity physics and sensor simulation large-scale parallel training and native ROS 2 interoperability-enabling an end-to-end workflow from virtual world construction to agent behaviour generation. Built on RaysEngine-a proprietary domestically developed rendering and computing foundation-the platform supports both on-premises and cloud deployment and is fully compatible with China The platform supports both on-premises and cloud deployment and is fully compatible with China's domestic software and hardware ecosystem ensuring technological autonomy and controllability. Core Capabilities Centered around The platform supports both on-premises and cloud deployment and is fully compatible with China “s domestic software and hardware ecosystem ensuring technological autonomy and controllability. RaysTwins delivers four integrated core capabilities: 1. Photorealistic Scene Rendering Leveraging advanced PBR++ shading and hybrid ray tracing RaysTwins unifies material appearance and This provides a high-fidelity visual environment essential for training robust robot perception and decision-making systems. 2. This provides a high-fidelity visual environment essential for training robust robot perception and decision-making systems. 2. Unified Physics Simulation The platform integrates multiple physics engines-including NVIDIA PhysX Bullet and MuJoCo-to support unified simulation of rigid bodies deformable objects (e.g. cloth) fluids and user-defined force fields. This multiphysics capability ensures numerical stability reproducibility and scalability for complex dynamic scenarios. 3. Seamless Sim-to-Real Integration With native ROS 2 compatibility RaysTwins enables synchronised time stepping topic mapping and data exchange between simulated and physical robots. It supports co-simulation verifiable sim-to-real transfer and rapid deployment-effectively bridging the gap between virtual development and real-world operation. It supports co-simulation verifiable sim-to-real transfer and rapid deployment-effectively bridging the gap between virtual development and real-world operation. 4. Scalable Parallel Training RaysTwins accelerates reinforcement learning and policy optimisation through multi-GPU support and edge edge support. RaysTwins accelerates reinforcement learning and policy optimisation through multi-GPU support and edge-cloud collaboration. It enables batched training and evaluation at scale while maintaining compatibility with It enables batched training and evaluation at scale while maintaining compatibility with domestic CPUs GPUs and operating systems-ensuring independent control over the entire AI development stack. Impact Beyond these core pillars RaysTwins incorporates generative scene modelling unified dynamics solving closed-loop virtual-real validation and distributed edge-cloud coordination. Together these features allow rapid construction of high-fidelity task-specific simulation environments. The platform delivers efficient reliable simulation support for a wide range of applications-including industrial robotic The platform delivers efficient and reliable simulation support for a wide range of applications-including industrial robotic manipulation service robot interaction autonomous mobile navigation and large-scale reinforcement learning. cycles reducing trial-and-error costs and enabling seamless transition from simulation to deployment RaysTwins accelerates the full innovation RaysTwins accelerates the full innovation lifecycle of intelligent embodied systems.