Wanxun launches NOVA 2.0, a flexible embodied brain trained on more than 10 million real-world operations

Wanxun Technology, a global leader in flexible embodied intelligence, has released NOVA 2.0 (Neuromorphic Omniversal Versatile Architecture), its second-generation flexible embodied brain and the industry’s first open-world, all-weather, commercial-delivery-grade embodied-intelligence architecture. It is built from more than 10 million multi-dimensional extreme-condition operations across 40-plus real industry scenarios, and its cross-scenario migration and delivery ability have won broad global customer recognition, forging a scarce real-world data flywheel.

Wanxun NOVA 2.0 flexible embodied brain demonstration
Wanxun’s NOVA 2.0 flexible embodied brain shown operating in a real scenario. (Source: Leiphone)

Embodied intelligence is moving from ideal-environment demos and early pilots toward open-world deployment, yet robots still face three challenges: long-chain reasoning latency against high-frequency response, severe scarcity of real multimodal operation data, and weak brain-body reliability in commercial delivery. Wanxun drew its first principle from 4 billion years of biological evolution: the natural coupling of flexible body and intelligence is the algorithm for physical-world randomness. Its ‘Somatic AI’ concept builds a flexible embodied architecture on soft-robot bodies.

NOVA 2.0 uses a tactile-native multimodal layered architecture with subconscious multimodal motion primitives for real-time trajectory generation and scene adaptation, giving three strengths: low data need, ultra-fast response and strong generalisation. Three migration breakthroughs stand out: one brain fits many forms, from muscle joints to soft single arm, soft dual arm and soft dexterous hand; over 1 billion dynamic parameter combinations across a 20-plus dimensional subconscious action space; and 20 millisecond trajectory generation that beats standard large models in sudden-condition correction.

Wanxun soft robotic arm handling a fragile object
A Wanxun flexible robotic arm performing delicate handling with tactile sensing. (Source: Leiphone)

The architecture has been applied across more than 40 real scenarios, including property construction, power energy, transport, smart driving and industrial manufacturing, with over 10 million operations. It works from zero lux to 100,000 lux, from minus 40 to 60 degrees Celsius, at 5,190 metres altitude, against 110 kilovolt electromagnetic interference, in 61 kilometre per hour wind and rain, and in Class 100 cleanrooms, needing as little as 1 per cent of the usual starting data.

Commercial results are growing: services across 40-plus industry scenarios with over 10 million operations, more than 90 per cent high-value orders, B-side satisfaction above 96 per cent and C-side above 97 per cent, repurchase orders above 40 per cent of value, over 50 Fortune 500 clients, and reach across more than 100 countries, regions and cities covering over 90 per cent of APEC economies.

Wanxun robot operating in an extreme environment
Wanxun’s system runs in extreme temperature, altitude and electromagnetic conditions. (Source: Leiphone)
Wanxun flexible robot in an industrial setting
A Wanxun flexible embodied robot deployed in industrial manufacturing. (Source: Leiphone)

From building tops and city canyons to high-voltage grids and snow plateaus, flexible embodied intelligence has crossed the open-world threshold into repeatable productivity. Driven by NOVA 2.0 and a data flywheel built from 100 per cent real data, Wanxun is extending the technology across more scenarios and robot forms.

Editor’s note: This is an adapted translation of the original Leiphone (Lingxiang Wang) report. It has been trimmed and restructured for readability for an international business audience. The full original (in Chinese) is at https://www.leiphone.com/category/robot/tEzDlRIgiHsXV8B0.html.

Translated and adapted from Leiphone (Lingxiang Wang) (https://www.leiphone.com/category/robot/tEzDlRIgiHsXV8B0.html).

Leave a comment