company blog about Supply Ambarella Automotive AI Solutions:CVflow® AI SoCs,Oculii™ AI Radar

Supply Ambarella Automotive AI Solutions:CVflow® AI SoCs,Oculii™ AI Radar

Shenzhen Mingjiada Electronics Co., Ltd. is a specialist distributor of electronic components, offering high-quality supply services backed by extensive procurement channels and in-depth procurement expertise.

Supply Advantages:

Brands and Certifications: As a globally recognised distributor of electronic components, we are certified to the ISO 9001:2014 Quality Management System and ISO 14001:2014 Environmental Management System standards.

Product Range and Inventory: With over 2 million SKUs in stock, our product portfolio covers microcontrollers (MCUs), field-programmable gate arrays (FPGAs), AI chips, sensors and power management ICs, meeting diverse ‘one-stop’ procurement needs.

Quality and Channels: We guarantee the supply of genuine products only. We have established partnerships with over 500 renowned global semiconductor manufacturers or their authorised channels, building a robust and reliable supply chain system, and conducting rigorous quality inspections prior to stock intake.

Delivery and Service: Leveraging our extensive inventory and global supply chain network, 90% of orders can be dispatched within 1–3 days. Urgent requests are responded to within 24 hours, and we offer global express delivery services with a 72-hour turnaround.

Expert Team: Our experienced professionals not only assist with product sales but also provide Bill of Materials (BOM) optimisation, component selection support and small-batch sample supply, helping customers reduce overall procurement costs.

Ambarella Automotive AI Solutions: CVflow® AI SoC and Oculii™ AI Radar, reshaping the core of intelligent driving perception

As automotive intelligence advances towards advanced autonomous driving (Level 2+ to Level 4), efficient, reliable and low-power environmental perception capabilities form the cornerstone of system safety. Leveraging its deep expertise in visual processing and AI computing, Ambarella has launched an integrated automotive AI solution centred on the CVflow® AI SoC as its computational core, with the Oculii™ AI Radar extending its perception capabilities. Through the deep integration of ‘vision and radar’, this integrated automotive AI solution provides vehicle manufacturers and Tier 1 suppliers with an intelligent driving perception platform covering all scenarios and operating conditions, redefining the standards for energy efficiency and reliability in in-vehicle AI perception.

I. CVflow® AI SoC: The Central Perception Brain of Intelligent Driving

Ambarella’s CVflow® series of automotive-grade AI chips are low-power, highly integrated, high-performance system-on-chip (SoC) solutions designed specifically for autonomous driving and Advanced Driver Assistance Systems (ADAS). Their core strength lies in the deep integration of leading-edge AI computing, professional-grade image processing, multi-sensor fusion and automotive-grade safety mechanisms, enabling a single chip to support the full stack of requirements from integrated driving and parking to domain-controlled autonomous driving.

1. Core Architecture: CVflow® AI Engine, Industry-Leading Energy Efficiency

The CVflow® architecture represents Ambarella’s core technological advantage, employing a heterogeneous computing design comprising a dedicated Neural Vector Processor (NVP) and a General-Purpose Vector Processor (GVP):

NVP (Neural Vector Processor): Optimised specifically for deep neural networks (DNNs), it supports real-time inference of complex models such as CNNs and Transformers. With a single-chip computing power of up to 500 TOPS (e.g., CV3-AD685), it can efficiently run algorithms for BEV perception, object detection and semantic segmentation.

GVP (General-Purpose Vector Processor): Responsible for traditional computer vision, radar signal processing and floating-point operations. It works in tandem with the NVP to significantly free up CPU resources, achieving industry-leading AI performance per watt.

Advanced Process Technology: The flagship CV3-AD series utilises a 5nm process, reducing power consumption by 40% compared to the previous generation, thereby meeting the stringent thermal and power constraints of in-vehicle domain control.

2. Full-stack Perception Capabilities: A Hardware Foundation for Multi-sensor Fusion

The CVflow® SoC incorporates powerful signal processing units and a rich array of interfaces, making it an ideal hardware platform for multi-sensor data fusion:

Top-tier ISP image processing: Supports 8K@60fps video encoding, multiple exposure HDR, 3D noise reduction and fisheye correction, delivering clear images even in harsh conditions such as 0.01 Lux ultra-low light or strong backlighting, providing a high-quality data source for AI perception.

Massive Sensor Connectivity: A single chip can simultaneously process data from 16 cameras, multiple millimetre-wave radars, LiDAR and IMUs, supporting low-latency synchronisation and spatio-temporal alignment.

High-Speed Communication Interfaces: Integrated multi-channel PCIe, CAN FD, Gigabit Ethernet and more, meeting the high-speed data exchange requirements with other ECUs in the vehicle under a domain-controlled architecture.

3. Automotive-grade security: Dual safeguards for functional and information security

The entire product range strictly adheres to the highest safety standards for automotive electronics, establishing a robust security defence for autonomous driving systems:

Functional safety: Certified to ISO 26262 ASIL-B/D, the CV3-AD series is equipped with three pairs of lockstep R52 safety islands and redundant designs for critical modules, ensuring safe responses in the event of system failures.

Cybersecurity: Integrated with a Hardware Security Module (HSM), it supports secure boot, memory encryption, TrustZone, TRNG (True Random Number Generator) and virtualisation isolation to defend against cyberattacks and data tampering.

4. Core Product Series: Covering all scenarios from L2 to L4

CV3-AD Series: Flagship AI domain controller chips (e.g. CV3-AD685), designed for Level 4 autonomous driving and central computing platforms, supporting full-scenario sensor fusion and complex decision-making algorithms.

CV22FS/CV20AQ: Mid-to-high-end ADAS chips, supporting L2+ applications such as forward-facing monocular/stereoscopic vision, Driver Monitoring System (DMS) and electronic rear-view mirrors, certified to ASIL C by Ambarella.

CV75A/CV5 Series: Entry-level to mid-range solutions, designed for applications such as dashcams, electronic rear-view mirrors and parking assistance, balancing cost and AI performance.

II. Oculii™ AI Radar: AI-driven 4D imaging millimetre-wave radar

Oculii, a leader in AI radar technology, has integrated its adaptive AI 4D imaging radar technology into automotive solutions. This addresses the shortcomings of pure vision systems in detecting targets under extreme weather conditions (rain, fog, dust storms), at long ranges, and with low reflectivity, forming an all-weather, multi-dimensional perception loop combining ‘vision and radar’.

1. Core Technology: Virtual Aperture Imaging, AI-Reconstructed Radar Performance

Oculii™ has completely revolutionised the traditional hardware approach of relying on “increasing antenna arrays” to improve resolution. Instead, it employs AI-driven adaptive waveform processing and virtual aperture imaging technology:

AI Adaptive Waveforms: Algorithms analyse the environment in real time to dynamically optimise radar transmission waveforms, boosting standard radar resolution by up to 100 times without the need for additional physical antennas.

Virtual Antenna Array: AI-driven sparsification algorithms create a “virtual antenna”, enabling performance on par with traditional high-resolution radars using just one-tenth the number of physical antennas, thereby significantly reducing hardware costs, size and power consumption.

4D High-Density Point Cloud: Outputs an ultra-dense radar point cloud containing tens of thousands of points per frame, precisely measuring the target’s distance, velocity, azimuth, and elevation (4D information) to reconstruct a three-dimensional contour of the target.

2. Breakthrough Performance: Long Range, High Accuracy, Low False Alarms

Extremely Long Detection Range: Maximum effective detection range exceeds 500 metres, capable of accurately identifying low-profile obstacles the size of a mineral water bottle from over 100 metres away.

Ultra-High Angular Resolution: Horizontal resolution of 0.5°, vertical resolution of 1°, enabling clear separation of small, close-range targets such as pedestrians, motorcycles and guardrails in congested traffic conditions.

Ultra-low false alarm rate: AI algorithms effectively suppress ‘ghosting’ and multipath interference. The system has passed high-speed AEB (Automatic Emergency Braking) tests conducted by leading global OEMs, achieving industry-leading recognition rates for easily missed targets such as moving children and pets.

All-weather operation: Unaffected by lighting conditions, rain, fog or dust, it perfectly complements cameras to ensure reliable perception even in extreme weather conditions.

3. Product Form Factors: Flexible Deployment, Suitable for All Scenarios

Software Licensing Solution: The Oculii™ AI radar algorithm can be directly deployed on Ambarella CVflow® SoCs (such as the CV3-AD series), working in conjunction with third-party RF chips to enable centralised radar signal processing and deep sensor fusion.

Integrated Radar Modules: Offers pre-integrated reference designs such as EAGLE and FALCON (e.g., the Eagle-C 77GHz radar head), providing plug-and-play functionality to shorten customer development cycles.

Low Power Consumption: Module power consumption is less than 2.5W, significantly lower than traditional high-resolution radars, making it suitable for automotive low-voltage power supply environments.

III. CVflow® + Oculii™: Fusion Perception, Setting a New Benchmark for Autonomous Driving

The core value of Ambarella’s solution lies in the deep synergy between the CVflow® AI SoC and the Oculii™ AI radar, rather than a simple aggregation of sensors:

Low-level data fusion: Low-level, high-bandwidth fusion of raw radar data and camera images is achieved within the CVflow® SoC, resulting in lower latency and higher accuracy compared to traditional approaches where sensors process data separately before fusion.

Optimised computing power sharing: Oculii™ radar algorithms and CV vision algorithms share the computing power of the CVflow® NVP/GVP, eliminating the need for a dedicated radar processing chip and reducing system BOM costs and complexity.

All-scenario perception redundancy: Vision handles detail recognition and semantic understanding; radar handles distance and speed measurement, providing all-weather coverage. The two complement each other’s strengths to build a seamless, highly reliable perception barrier, meeting the stringent safety redundancy requirements of Level 4 autonomous driving.

Unified Development Platform: Provides an integrated SDK and toolchain, supporting rapid algorithm porting, debugging and deployment, helping OEMs and Tier 1 suppliers shorten development cycles and accelerate the mass production of advanced autonomous driving solutions.

IV. Application Scenarios: Empowering Full-Stack Intelligent Driving from L2+ to L4

Advanced ADAS (L2+/L3): Supports functions such as Highway NOA (Navigation-Assisted Driving), Urban NGP, automatic lane changing, traffic jam cruise control and fully automated parking, addressing complex urban scenarios and unexpected situations.

Automatic Emergency Braking (AEB): Precisely identifies pedestrians, non-motorised vehicles and crossing obstacles, significantly reducing the risk of collisions.

Driver/Occupant Monitoring (DMS/OMS): Combines visual AI to enable fatigue detection, distraction alerts and occupant status monitoring.

Electronic Rear-view Mirror (CMS): High-definition video processing + AI blind-spot detection, replacing traditional optical rear-view mirrors to enhance visibility and safety.

Conclusion

Ambarella’s integrated automotive AI solution, comprising the CVflow® AI SoC and Oculii™ AI radar, provides a comprehensive hardware and software foundation for intelligent driving—from perception and fusion to decision-making—through leading energy efficiency, deep sensor fusion, and automotive-grade safety and reliability. At this critical juncture where autonomous driving technology is transitioning from ‘early adoption’ to ‘widespread adoption’, Ambarella is driving the leapfrog development of the smart mobility industry through continuous technological innovation, helping global automotive manufacturers build safer, smarter and more efficient next-generation smart vehicles.