Embodied AI & Cross-Scenario Task Management:
Powered by the self-developed Robot Operator Model-1 (ROM-1), featuring 100M parameters and 24Hz end-to-end control capability. Through imitation and reinforcement learning, ATOM autonomously decomposes complex tasks and makes decisions, seamlessly adapting to diverse commercial and industrial environments such as automotive assembly line material handling, beverage preparation in coffee shops, and pharmacy nighttime dispensing.
Neuro-Driven Dexterous Manipulation:
Features the Neuro-Driven Dexterity System (NDS) combined with a Transformer-based architecture and binocular RGB vision, precisely controlling 28 upper-body degrees of freedom (DoF). With servo-level vibration suppression and 200Hz high-frequency control, it exhibits smooth, human-like fine motor skills to effortlessly handle challenging tasks such as tool use, precision assembly, and multi-specification material handling.
Anthropomorphic Straight-Knee Walking & High Efficiency:
Utilizes the advanced Anthropomorphic Walking System (AWS) that deeply mimics human biomechanics. Compared to traditional bent-knee walking robots, ATOM's straight-knee walking technology reduces energy consumption by 42%, significantly enhancing stability and endurance for continuous operations, allowing it to flexibly navigate narrow spaces and perfectly adapt to various workstation heights from 700mm to 1000mm.
Powerful Edge Computing & Safe Human-Robot Collaboration:
Equipped with an advanced edge computing system that delivers 7.7 times the industry-standard computational power. Integrated with human-robot interaction and safety technologies, it enables ATOM to seamlessly coordinate with human workers or other robots, maximizing efficiency in highly synchronized production lines or multi-device service scenarios.
Dedicated Research Models & Open AI Ecosystem:
To meet advanced scientific research and large-scale data collection needs, specialized models are available, including the ATOM Max (featuring 41 DoF, 1500 TOPS AI module, and VR/MR teleoperation) and the ATOM D (optimized for data collection). Providing high-precision URDF models, open-source training frameworks, and multi-modal data tools, it significantly lowers the barrier to entry for embodied AI and automation innovation.









