FIVEAGES Technical Overview ========================= Empowering Millions of Robots To Serve Humanity Let Human Being Be Human FIVEAGES (FiveAges Intelligent Technology Co., Ltd.) is an embodied intelligence pioneer delivering universally generalizable embodied brains and humanoid robots. This document provides a high-level overview of our platform architecture, foundation models, and deployment capabilities. ================================================================================ PLATFORM OVERVIEW ================================================================================ FIVEAGES delivers end-to-end embodied intelligence through an integrated technology stack: - Embodied foundation model series (FAM Series) for ultra-low-shot manipulation - Next-Generation Embodied World Model (BridgeV2W) for predictive foresight - Human-in-the-Loop Reinforcement Learning for continual adaptation - Self-developed wheeled humanoid robot platform (FA-L Series) - High-performance on-robot edge computing and safety control separation - Global deployment services and open brain hardware manufacturer ecosystem ================================================================================ FOUNDATION LARGE MODELS (EMBODIED BRAIN) ================================================================================ FAM Series (Ultra-Low-Shot Manipulation Large Model) --------------------------------------------------- World’s first ultra-low-shot end-to-end embodied manipulation large model. FAM preserves spatial structure across perception, language, and action through a unique "heatmap alignment" architecture, eliminating the spatial information loss common in traditional 1D latent vector representations. Key capabilities: - Lossless 2D and 3D spatial heatmap representations - 3–5 demonstrations for typical real-world task adaptation - Rapid cross-task and cross-environment policy transfer via data flywheel BridgeV2W (Next-Generation Embodied World Model) ---------------------------------------------- Transforms robot actions into pixel-level motion video futures given current observations and candidate action sequences. Enables physical robots with predictive capabilities and autonomous obstacle avoidance before physical execution. Key capabilities: - High-fidelity future visual state prediction in multi-view and cross-platform setups - Candidate action scoring and autonomous collision risk evaluation - State-of-the-Art benchmark performance under unseen viewpoints and complex scenes Human-in-the-Loop Reinforcement Learning ---------------------------------------- A closed-loop evolutionary paradigm breaking through scene generalization bottlenecks: - Few-shot diffusion visuomotor policy pretraining - Real-world error recovery intervention to eliminate error accumulation - Autonomous process and outcome reinforcement learning optimization - Multimodal human feedback alignment for preference adherence ================================================================================ TECHHIGHLIGHTS & DATA TECHNOLOGY ================================================================================ Robot-Centric Visual Flow -------------------------- Extracts task-relevant hand, object, motion, and occlusion relationships from human operation videos to accelerate policy training. Task-Centric 3D Virtual View for VLA ------------------------------------- Synthesizes task-relevant virtual camera views from multiple visual inputs, reducing viewpoint dependency without requiring explicit 3D rendering. Cross-View Data Augmentation ----------------------------- Expands demonstrations across viewpoints through motion retargeting and generative video completion, reducing manual multi-angle data collection. Embodied Visual Information Enhancement --------------------------------------- Detects uncertainty and preserves task-relevant visual evidence over long-horizon tasks through active attention and sensory fusion. ================================================================================ ROBOT PLATFORM ================================================================================ FIVEAGES FA-L Series -------------------- A self-developed wheeled humanoid platform designed for industrial and commercial operations. Key specifications: - Full-body degrees of freedom: 28 DoF (including 2 lumbar + 2 leg DoF adaptive lifting) - Dual 7-DoF humanoid robotic arms with sub-millimeter precision force control - Payload: >= 5 kg rated (single arm), 20 kg maximum (dual arm) - 4-wheel steering & 4-wheel drive omnidirectional mobile chassis (1.5 m/s speed) - Dual 3D LiDAR full-range sensing without blind spots & sensitive collision detection - Dual hot-swap batteries for continuous 24/7 uninterrupted operation - Secondary development API and OTA firmware/software evolution - Open interfaces for ERP, MES, and WMS integration ================================================================================ GLOBAL SERVICE & DEPLOYMENT CAPABILITY ================================================================================ Safety Architecture ------------------- Safety control is strictly separated from model inference. A dedicated safety control unit enforces physical limits, certified deceleration rules, and emergency behaviors. Enterprise Integration ---------------------- Open interfaces for ERP, MES, and WMS integration enable robots to receive tasks directly from existing operational processes. Global Service Capability ------------------------- Universal embodied brain delivery with global deployment readiness, standardized lifecycle maintenance, and customer-specified local inference data policies. ================================================================================ ABOUT FIVEAGES ================================================================================ FiveAges Intelligent Technology Co., Ltd., founded in September 2024, inherits decades of pioneering academic research from CASIA (Institute of Automation, Chinese Academy of Sciences) and Tsinghua University. - 10+ Champions in top global AI and robotics competitions (ICRA, IROS, CVPR) - 200+ Team Members, 85% R&D personnel, 1,000+ core papers, 40,000+ citations - Backed by leading institutional shareholders: Sinoking Capital, Oriental Fortune Capital, Cowin Capital, CAS Capital, HongShan China, CDH Investments, Fortera Capital, SinoWisdom Headquarters: 5F, Satellite Building, Haidian District, Beijing Business & Deployment: contact@five-ages.com Marketing & Media: marketing@five-ages.com Website: https://www.five-ages.com ================================================================================ © 2026 FiveAges Intelligent Technology Co., Ltd. All rights reserved.