Zhuoyu Technology

"Chengxing" is ZYT's flagship mass-production assisted driving product, branded internationally as ClixPilot. Built on the Chengxing Platform's 32-1500+ TOPS of AI compute and flexible computing platforms including cockpit-driving integration, it pairs a diverse sensor set with a highly perceptive, reasoning end-to-end algorithm architecture to deliver production-ready assisted driving features such as urban navigation, parking-space-to-parking-space driving and cross-level memory parking, while an end-to-end world model enables a generative driving experience with natural language interaction.

Compute range from 32 TOPS to 1500+ TOPS; sensor configurations supporting 7V/10V/12V/10V1L/12V1L; urban and highway navigation achieved with only 32 TOPS, with 100 TOPS unlocking additional features; automotive-grade safety operating system software compliant with ISO 26262 functional safety, with production software independent of open-source code; supports battery-electric, hybrid and pure internal combustion vehicles (equal intelligence across powertrains).

• Mass production of L2 and higher advanced assisted driving for passenger cars
• Urban navigation and highway navigation
• Parking-space-to-parking-space and cross-level memory parking
• Intelligentization of combustion-engine vehicles (equal intelligence across powertrains)
• Assisted driving for commercial vehicles and heavy trucks

The Chengxing Platform is ZYT's unified hardware-software technology foundation for mass-produced assisted driving, comprising a flexible-compute computing platform, diverse sensor configurations and an end-to-end algorithm architecture. The platform is highly integrated, cost-effective, highly modular and highly safe, adapting flexibly to different customer use cases and fitting into passenger cars of different powertrain types under power and resource constraints.

Hardware options include 7V mid-compute (32 TOPS/96 eTOPS), 7V/10V high-compute (100 TOPS/200 sparse TOPS), flagship compute (700/1000 TOPS) and cockpit-driving integrated combinations; expandable with the JIMU system, third-party LiDAR and maps; supports add-on innovations such as road preview, smart headlights and in-vehicle drones; compliant with ISO 26262 functional safety and ISO 21434 cybersecurity standards.

• Hardware carrier for mass-produced passenger car assisted driving
• Platform-level adaptation across vehicle models and price segments
• Unified hardware adaptation across verticals (lowering adaptation cost)
• Deployment of assisted driving functions and visualization systems

ZYT's highly perceptive end-to-end large model algorithm differs from mechanical imitation learning that behaves like a human; it aims for thinking that is like a human, with the perceptiveness to reason autonomously. Through train-infer-insight technology the model can generalize from a single example and grasp a point instantly, while 3D spatial attention and a safe reasoning framework make output trajectories safer and explainable. It is highly adaptable and can be deployed flexibly across sensor configurations, with or without LiDAR, and on different compute platforms.

Supports 32-1000 TOPS compute and 7V/10V/12V multi-sensor configurations; stereo-enhanced end-to-end architecture delivering 10-20x higher data training efficiency than conventional monocular or heterogeneous schemes; cross-configuration and cross-platform training plus a safe reasoning framework; production software independent of open-source code; backed by a large-scale automated data closed-loop toolchain supporting 4D ground-truth production pipelines, cloud simulation testing and OTA updates.

• End-to-end algorithms for urban navigation and highway navigation
• Long-tail scenarios and complex intersections
• Personalized driving style learning and imitation
• Shared model foundation for continuous evolution from L2 to L4

GenDrive is a new-generation assisted driving capability that ZYT launched first in the industry based on its end-to-end world model, and the industry's first personalized generative driving experience supporting natural language interaction. It delivers a driving style tailored to each individual and lets the assisted driving system understand user requests, aligning in real time with the driver's intent.

Based on an end-to-end world model architecture; supports VLA driving style control, VLA natural language interaction and VLA vision-language navigation; includes style, vision, navigation, trajectory and text encoders; adopts pre-training (IL) and post-training (IL+RL) paradigms.

• Personalized driving style learning and imitation
• Natural language and voice interaction with assisted driving
• Generative driving experience

ZYT's assisted driving controller is an automotive-grade, high-performance, highly scalable general-purpose computing platform designed specifically for assisted driving systems. Developed strictly in line with ISO 21434 cybersecurity standards and ISO 26262 functional safety requirements, it meets ASIL-D design standards and allows compute to be configured flexibly according to system needs and each customer's actual situation.

Configurable compute from 32 TOPS to 1500+ TOPS (the domain controller series as a whole covers 200-1400 TOPS mid-to-high compute); meets ASIL-D design standards; compliant with ISO 26262 functional safety and ISO 21434 cybersecurity requirements; IP6K7 ingress protection; meets the industry's highest reliability testing standards; highly integrated core components including safety MCU, ISP, GNSS and ultrasonic control units.

• Mass-produced passenger cars with mid-compute assisted driving
• Carrier for urban navigation and highway navigation functions
• Domain controllers for heavy trucks and commercial vehicles

The air-cooled high-compute intelligent driving domain controller is the air-cooled variant in ZYT's domain controller line, aimed at vehicles with high compute demand. It uses a unified architecture to carry the high-compute platform and works with diverse sensor configurations to support advanced assisted driving functions such as urban navigation, fitting vehicle platforms that lack liquid-cooling provisions.

Part of ZYT's 200-1400 TOPS mid-to-high compute domain controller series; air-cooled thermal design; compliant with ISO 26262 functional safety and ISO 21434 cybersecurity standards, meeting ASIL-D design standards and IP6K7 ingress protection.

• Mass production of urban navigation in high-compute passenger cars
• High-compute assisted driving for combustion-engine and mild-hybrid models
• Vehicle platforms without liquid-cooling provisions

The liquid-cooled high-compute intelligent driving domain controller is the liquid-cooled variant in ZYT's domain controller line, aimed at scenarios with prolonged high-load operation. Leveraging a highly modular design, it carries high-compute assisted driving algorithms while improving heat dissipation efficiency and sustained performance stability.

Part of ZYT's 200-1400 TOPS mid-to-high compute domain controller series; liquid-cooled thermal design; compliant with ISO 26262 functional safety and ISO 21434 cybersecurity standards, meeting ASIL-D design standards and IP6K7 ingress protection.

• Mass production of urban navigation in high-compute passenger cars
• Prolonged high-load assisted driving operation
• Assisted driving for commercial vehicles and trunk-line logistics

The cockpit-driving integrated intelligent driving domain controller is a single-chip cockpit-driving fusion computing platform jointly introduced by ZYT and Qualcomm. It delivers end-to-end urban NOA under a single-chip cockpit-driving integrated scheme, featuring more efficient cockpit-to-driving coordination, a simplified vehicle architecture and system-level safety restructuring, and it brings VLA into a unified architecture.

Based on Qualcomm SA8775P / SA8797 platforms; single-chip cockpit-driving integrated architecture; supports end-to-end urban NOA (industry-first in October 2025); compliant with ISO 26262 functional safety and ISO 21434 cybersecurity standards.

• Mass-produced passenger cars with cockpit-driving fusion
• Single-chip cockpit-driving integrated urban NOA
• Simplification of vehicle E/E architecture and cost optimization

The flagship-compute intelligent driving domain controller is ZYT's compute platform for advanced L3/L4 autonomous driving. It carries two NVIDIA Thor chips to form a triple-redundant controller and works with the in-house JIMU forward perception system and Zhizhou blind-spot-filling radar, purpose-built for scenarios such as Robotaxi.

Based on dual NVIDIA Thor chips; triple-redundant safety architecture; compute spanning 700/1000 TOPS up to the 1400 TOPS class; compliant with ISO 26262 functional safety and ISO 21434 cybersecurity standards; meets ASIL-D design standards.

• Robotaxi autonomous driving
• Redundant architecture for mass-produced L3/L4 autonomous driving
• Driverless logistics vehicles

ZYT's proprietary inertial navigation stereo binocular camera not only recognizes common targets on the road accurately, but also measures distance in three-dimensional space through the parallax principle, obtaining 3D position information about objects. It detects obstacles of any type and stands out for safety and reliability. As the core sensor for L2+ assisted driving, it supports assisted driving functions including urban navigation.

Three-dimensional ranging based on the parallax principle, providing high-precision distance measurement and high-resolution dense 3D point clouds; stereo-enhanced end-to-end architecture delivering 10-20x higher data training efficiency than conventional monocular or heterogeneous stereo schemes; joint optimization across camera module design, calibration, sensor layout and integration allows flexible configuration of millimeter-wave radar, ultrasonic sensors and LiDAR; core sensors developed and manufactured in-house.

• Core perception for L2+ passenger car assisted driving
• Urban navigation
• Detection of any obstacle type and intelligent collision avoidance
• Innovative functions such as road preview and smart headlights

The inertial navigation stereo trinocular camera combines an inertial navigation stereo binocular unit with a telephoto camera system, delivering high-quality 3D point clouds together with long-range perception. Powered by ZYT's Stereo Occupancy Network technology, it detects ordinary static and dynamic obstacles and estimates their position and future motion, while the telephoto camera more than doubles the detection range for traffic lights and small obstacles.

Composed of bionic stereo vision plus a telephoto monocular camera; uses Stereo Occupancy Network technology; the telephoto camera more than doubles the detection range for traffic lights and small obstacles; core sensors developed and manufactured in-house.

• Advanced passenger car assisted driving (e.g. Hongqi Tiangong 08)
• Long-range perception and handling of complex dynamic scenes
• L3/L4 redundant perception architecture

The monocular vision camera and long- and short-focus vision modules in ZYT's perception hardware matrix build a multi-tier visual perception system, supplying forward and long-range visual input to assisted driving systems. Combined with the inertial navigation binocular and trinocular units, the long- and short-focus vision modules complete a full visual perception product line-up.

Includes monocular vision cameras and long- and short-focus vision modules; together with the inertial navigation binocular and trinocular units it forms a complete perception hardware matrix; supports sensor configurations such as 7V/10V/12V/10V1L/12V1L; core sensors developed and manufactured in-house.

• Visual perception for passenger car assisted driving
• Long-range target and traffic light recognition
• AVM surround view and panoramic DVR output

JIMU 2.0 is an automotive-grade, highly integrated all-in-one sensor combining a wide-angle camera, a telephoto camera and LiDAR in a smaller package with stronger performance. The LiDAR intelligently adjusts its field of view, angular resolution and detection range according to the current scene. The system can be installed inside the cabin or have its LiDAR mounted separately on the roof, and it is ZYT's original LiDAR-vision front-fusion solution.

Automotive-grade highly integrated all-in-one design combining wide-angle camera, telephoto camera and LiDAR; the LiDAR intelligently adjusts field of view, angular resolution and detection range by scene; supports in-cabin installation or separate roof mounting of the LiDAR; meets the redundancy architecture design requirements of L3/L4 autonomous driving systems.

• Advanced assisted driving in complex urban scenarios
• Preparation for mass production of L3/L4 autonomous driving
• Commercial heavy truck and bus scenarios
• Enhanced perception under extreme lighting and heavy traffic

Zhizhou is an automotive-grade, fully solid-state, ultra-wide-angle blind-spot-filling LiDAR with a small footprint, high reliability and high performance, significantly improving safety in driving and parking scenarios. Designed specifically for L3/L4 autonomous driving systems, it can be mounted on the left and right fenders and the front and rear bumpers for easy integration.

Automotive-grade fully solid-state design; ultra-wide field of view of H160° x V144°; compact size, high reliability and high performance; designed specifically for L3/L4 autonomous driving systems; mountable on left and right fenders and front and rear bumpers for easy integration.

• Near-range blind-spot perception for L3/L4 autonomous driving
• Improved safety in driving and parking scenarios
• Near-range perception integration on fenders and bumpers
• Robotaxi and driverless logistics vehicles

The integrated binocular + LiDAR perception module is an all-in-one perception product in ZYT's perception hardware matrix that combines an inertial navigation stereo binocular unit with LiDAR. Through a LiDAR-vision front-fusion approach it improves perception stability in complex scenarios and forms an important part of ZYT's unified automotive-grade hardware matrix.

Integrates an inertial navigation stereo binocular unit with a LiDAR module; uses a LiDAR-vision front-fusion approach; automotive-grade manufacturing produced as a one-stop service by ZYT's intelligent manufacturing centers (dual bases in Shenzhen and Changzhou), covering vision sensors, controllers and core LiDAR components.

• Forward perception for advanced assisted driving
• Enhanced perception stability in complex scenarios
• Cross-vertical platforms for commercial and passenger vehicles

The mass-production assisted driving feature suite offered by ZYT Chengxing spans the complete range from active safety and parking assist to highway and urban navigation. Active safety covers FCW, AEB, LDW, LDP, BSD and DOW; parking assist includes assisted parking, smart summon and remote parking; memory parking supports indoor and outdoor car parks with designated, zone and exploratory modes; highway and urban navigation are both free of high-definition map dependency and work anywhere in the country.

Active safety includes FCW, AEB, LDW, LDP, BSD and DOW; parking assist includes assisted parking, smart summon and remote parking; memory parking supports indoor and outdoor car parks and three modes (designated, zone, exploratory); cross-level memory parking works upward and downward across levels, covering the full scenario from the toll barrier to the parking space; highway navigation includes toll booth passage, intelligent speed control and automatic on/off ramps, with no HD map dependency; urban memory navigation needs a route to be memorized only once, with no cloud interaction; urban navigation handles construction zones, overtaking and detours, and intersection passage.

• Active safety and passive protection for passenger cars
• Integrated driving-and-parking scenarios
• Navigation on highways and urban expressways
• Navigation on urban roads

A collection of vehicle-level innovative applications built on the perception and computing capabilities of ZYT Chengxing. Road preview uses terrain recognition and road elevation detection to support intelligent adjustment of driving modes and suspension parameters; smart headlights use binocular ambient brightness measurement plus vehicle and VRU detection for automatic on/off and intelligent control of lighting zones; panoramic DVR outputs all-around camera video streams; automatic wipers recognize rainfall intensity visually and adjust the setting automatically; fog detection identifies windshield fogging and activates defogging automatically; anti-theft monitoring uses all-around cameras to detect scratches and theft and pushes alerts to the head unit.

Road preview includes terrain recognition and road elevation detection; smart headlights are based on binocular ambient brightness measurement and vehicle/VRU detection; panoramic DVR outputs all-around camera video streams to the head unit; automatic wipers can eliminate the vehicle's rain sensor; fog detection automatically activates the defogging function; anti-theft monitoring uses all-around cameras for anomaly detection and sends events to the head unit.

• Intelligent chassis and suspension adjustment for passenger cars
• Intelligent lighting control
• Driving recording and panoramic imaging
• Automated wipers and defogging
• Vehicle-level intelligent anti-theft monitoring

A highway NOA product that ZYT has introduced for the commercial heavy truck sector. It already covers all of the top six domestic brands and has partnered with the European brand Scania. The first highway NOA product has entered mass production and is the first product in China's commercial vehicle industry to satisfy the mandatory national standard for combined driving assistance.

Based on the Chengxing Platform's unified automotive-grade hardware and software architecture; aimed at highway scenarios in trunk-line logistics; domain controllers are brought into production at the Changzhou manufacturing base; the first heavy truck models are scheduled for official mass production in the first half of 2026.

• Highway navigation for trunk-line logistics heavy trucks
• Reducing driver fatigue on long heavy truck journeys
• Improving safety and efficiency in trunk-line logistics

ZYT has teamed up with Yutong to develop a commercial bus NOA product, scheduled for mass-production delivery in September 2026 and to cover more bus categories in 2027. It is an important business line for ZYT in the commercial vehicle vertical.

Based on the Chengxing Platform's unified automotive-grade hardware and algorithm foundation; jointly developed with Yutong; mass-production delivery planned for September 2026, expanding to more bus categories in 2027.

• Intelligent assisted driving for commercial buses
• Mass-production delivery of NOA for passenger transport scenarios

ZYT is working with leading commercial vehicle companies to design and define driverless logistics vehicles for the intelligent transformation of scenarios such as mines and ports. In this project ZYT acts not merely as a Tier 1 supplier but takes part in product design and many other stages, with the business progressing on plan toward mass-production introduction and trial operation.

Based on unified automotive-grade hardware and the native multimodal foundation model; adopts a continuous L2-to-L4 evolution architecture sharing one foundation model and adding stricter ODD detection, redundant safety degradation and a back-office operations system; the Changzhou plant initially introduces production of the domain controller for one driverless logistics vehicle product.

• Driverless logistics in mines
• Driverless logistics in ports
• Automated delivery in industrial parks

ZYT's L4 autonomous driving solution for the Robotaxi direction, purpose-built for that scenario around a triple-redundant L4 controller based on dual Thor chips, together with the in-house JIMU forward perception system and Zhizhou blind-spot-filling radar. The business is currently progressing on plan toward mass-production introduction and trial operation.

Carries two NVIDIA Thor chips to form a triple-redundant controller; paired with the in-house JIMU forward perception system and Zhizhou fully solid-state blind-spot-filling LiDAR; follows a continuous L2-to-L4 evolution path sharing one native multimodal foundation model, with L4 adding stricter ODD detection, redundant safety degradation and a back-office operations system.

• Robotaxi driverless mobility
• Trial operation of L4 autonomous driving

Guided by the vision of building a spatial intelligent mobility foundation and leading the era of autonomous mobile robots, ZYT is extending the boundaries of mobile intelligence from passenger car assisted driving into a broader range of business scenarios, providing underlying technology for autonomous mobile robots in different settings and enabling an ecosystem of intelligent mobility from the ground to near-ground space. The native multimodal foundation model supports six verticals: passenger cars, commercial heavy trucks, buses, driverless logistics vehicles, Robotaxi and in-vehicle drones.

Built on a data-driven development paradigm, a mature foundation model and integrated hardware-software engineering capability; the native multimodal foundation model supports six verticals; a unified automotive-grade hardware matrix greatly reduces cross-vertical adaptation cost; the Changzhou plant is planned to take on production of products such as in-vehicle drones.

• Underlying technology for autonomous mobile robots
• Embodied intelligence
• In-vehicle drones
• All-domain intelligent mobility ecosystem