Sidi Moves Towards "Heavy-Duty Embodied AI", AI's Impact on the Physical World Gets a New Interpretation

As the global AI wave focuses on algorithm iterations in the digital world, a revolution about reconstructing productivity in the physical world is quietly occurring.

At the 2026 Asia Logistics Biennial, Sidi Intelligent Driving held a press conference, systematically showcasing core achievements towards "Heavy-Duty Embodied AI", launched the industry's first V2P (Vehicle-to-Pedestrian) portable intelligent terminal TAG, and signed cooperation agreements with Proton Auto and Deep Robotics.

Let Machines Go to the Most Dangerous Places

Every productivity leap in human civilization begins with the emergence of a new tool. The steam engine brought industrialized production, the internal combustion engine promoted globalization, and computers opened the information age. From physical liberation to power upgrade, then to computing power revolution, the essence of every leap is the extension of human capability boundaries.

Standing at a new node, what is the next tool to change the physical world?

Sidi Intelligent Driving's answer is: Heavy-Duty Embodied AI.

Heavy-Duty Embodied AI refers to endowing heavy machinery with autonomous operation capabilities through intelligent technology. Heavy machinery features high tonnage, high load capacity, and adaptability to extreme operating conditions. They are used in scenarios such as breaking walls, drilling holes, explosives loading, excavation, loading, and heavy-duty transportation, making them the hardest-core production tools. Physical AI gives heavy machinery a complete perception system, decision-making brain, and execution capability, enabling independent operation in extremely complex production environments, and collaborating with other machines to complete comprehensive scenario tasks.

Understand with a formula: Heavy-Duty Embodied AI = Physical AI × Heavy Machinery × High-Risk Operation.

Physical AI is the technical engine, heavy machinery is the physical carrier, and high-risk operation is the application scenario. The three have a multiplier effect on each other, amplifying value.

Behind this direction lies Sidi Intelligent Driving's deep robotics genes. Led by founder Prof. Zexiang Li, the team started from rigorous mathematical and physical logic, built a systematic robotics knowledge system, and related monographs have become important theoretical references in the global robotics field, serving as crucial underlying support and innovative soul for the company's technology R&D and product development.

Relying on this technical accumulation, Sidi Intelligent Driving advances the Heavy-Duty Embodied AI layout along three lines: Heavy-Duty Autonomous Driving (Transport Robots), relying on autonomous mining trucks and autonomous heavy-duty trucks to overcome safety and efficiency bottlenecks in extreme operating conditions; Heavy Equipment Embodied AI (Operation Robots), multi-type embodied agents covering full-scenario operations from ground to underground; Intelligent Transportation, taking vehicle-spec high-performance perception technology as the base, covering multi-scenario safety and intelligent upgrade needs.

"Autonomous driving is essentially an autonomous mobile wheeled robot. Sidi Intelligent Driving has accumulated robotics core technologies and data such as heavy-duty control, extreme condition perception, and cluster scheduling. From wheeled robots to excavation, rock drilling, and explosives loading robots, it is a natural extension of the same technical vein." Dr. Hu Sibuo, CEO of Sidi Intelligent Driving, stated at the press conference.

▲ Dr. Hu Sibuo, CEO of Sidi Intelligent Driving, delivers the keynote speech

This innovative technology is continuously being validated in real business scenarios. Currently, Sidi Intelligent Driving's various intelligent products have been scaled and applied in resource mining and transportation, logistics, large infrastructure, and traffic engineering fields, with total shipments exceeding 20,000 units. By substituting or empowering high-risk and arduous manual operations, it continuously releases the deep efficiency of industrial and mining operations, driving the intergenerational leap of heavy-duty productivity.

In FY2025, Sidi Intelligent Driving's total revenue increased by 115.8% year-over-year, reaching 885 million yuan, gross profit increased by 86.7% year-over-year, accelerating towards the profitability inflection point.

▲ Sidi Intelligent Driving's unmanned container truck passes through the Zhongyue Puzhai-Xinqing Freight Dedicated Channel Checkpoint

Becoming the "Brain" of Heavy-Duty Embodied AI

In terms of business model, Sidi Intelligent Driving is positioned as the "Brain" of Heavy-Duty Embodied AI, focusing on technology empowerment, becoming the most trusted technical partner in the industry.

Based on this positioning, Sidi Intelligent Driving has built the "Heavy-Duty Embodied AI Technology Platform", divided into three-layer architecture.

The bottom layer is a unified data engine. Based on production, environmental, and vehicle data from various scenarios, it builds a standardized data base.

The multimodal perception fusion system can maintain continuous detection capability of dynamic and static obstacles in the whole domain under harsh operating conditions such as dust, vibration, extreme weather, and lighting changes. Its multi-source heterogeneous environmental data continuously enriches the scenario coverage and model training sample quality of the data base.

The middle layer is the heavy-duty world model. Adopting an edge-cloud collaborative architecture, terminal action models make millisecond-level real-time decisions, while cloud training models continuously iterate and evolve.

Different from other types of world models, the heavy-duty world model deeply integrates exclusive data of production operation scenarios, heavy-duty robot dynamics, and action space reasoning, and is native-designed for multi-agent collaboration. Its core capability lies in understanding the bidirectional interaction between agents and the physical environment, inferring the cumulative effect of the agent's own behavior on the environment in real-time, and making precise planning based on the evolved environment state. This closed-loop decision-making capability of environment evolution reasoning constitutes a unique technical barrier for heavy-duty scenarios.

The top layer is the cluster decision and scheduling system. Taking global optimality as the goal, it commands multi-type robots to collaborate in the same space-time domain. This highly mature multi-type cluster intelligent solution has already completed engineering validation first in mining scenarios where agents are densely mixed.

The advantage of the whole architecture lies in extremely strong generalization capability: the same system, multi-form adaptation, cross-scenario reuse. Model capabilities accumulated in a single scenario can achieve low-cost cross-scenario migration, realizing the continuous release of technical value.

At the press conference, Sidi Intelligent Driving also signed cooperation agreements with Proton Auto and Deep Robotics respectively.

Proton Auto is a complete vehicle enterprise under Shaanxi Auto Group focusing on new energy intelligent connected commercial vehicles, focusing on commercial vehicle intelligence, electrification, connectivity, and lightweighting. Sidi Intelligent Driving reached deep cooperation with Proton Auto. Both sides will take the autonomous heavy-duty truck overall solution as the core and jointly promote the application and promotion of autonomous driving technology in commercial vehicle scenarios.

Deep Robotics focuses on the R&D and industrialization of quadruped robots and human-like robots, and has achieved commercial landing in scenarios such as power supply, firefighting, and security. Both sides' cooperation will focus on mining robots.

Cooperation with Proton Auto and Deep Robotics is the concentrated manifestation of Sidi Intelligent Driving's ecosystem collaboration logic. From signing with Sinomine Mining and Sinomine Mine Research Institute in September 2025, to signing a 500 autonomous mining truck procurement order with Guangna Group in February 2026, joining hands with UK MMD Group and Schaeffler in April, to reaching cooperation successively with Yipule, Zoomlion, Tongyong Consulting, Proton Auto, and Deep Robotics in June, Sidi Intelligent Driving's "friend circle" is rapidly expanding along three clues: "carrier — scenario — going global".

Focusing on being the "Brain" of Heavy-Duty Embodied AI — this open positioning is attracting more and more industrial chain partners to join.

Departing from China, Moving Towards Global Layout

Heavy-Duty Embodied AI is a trillion-dollar market driven by policy and market. On the policy side, "no people means safety, fewer people means safety" is already a clear development direction, and the safety baseline is being converted into a top-level push for industrial upgrading. On the market side, AI needs computing power, and computing power needs energy. Heavy-Duty Embodied AI happens to stand at the intersection of AI revolution and energy infrastructure, combining the explosive power of tech growth with the certainty of the energy industry.

The time window has opened. Taking the largest commercial landing scenario mining as an example, the penetration rate of autonomous mining trucks is currently around 10%, and the inflection point for large-scale explosion is approaching. China's mature new energy vehicle supply chain is fully empowering embodied intelligence, allowing Sidi Intelligent Driving to step on the dividends of industrial acceleration.

A grander picture lies in going global. Sidi Intelligent Driving, as a Chinese enterprise that has completed commercial validation first, is taking the advantages of Chinese Heavy-Duty Embodied AI to the world.

Currently, Sidi Intelligent Driving has started global layout around resource-rich countries such as Australia, Brazil, and Indonesia, established cooperation with multiple mine owners, OEMs, and operators, and plans to gradually cover markets such as Africa, Central Asia, and the Middle East in the coming years, achieving commercial batch delivery, and establishing a leading position in overseas heavy-duty scenarios.

As new quality productivity, Heavy-Duty Embodied AI changes the most fundamental production mode of the physical world. The scenarios chosen by Sidi Intelligent Driving have common characteristics of high risk, arduousness, and high value, and possess deep professional barriers. Sidi Intelligent Driving will liberate humans from dangerous and arduous operation scenarios with systemic innovation, reshape the physical world with sustainable intelligent transportation power, and promote the leap of industrial civilization paradigms.