Rapid Prototyping is an efficient product development strategy, especially in the field of autonomous vehicles, which significantly accelerates the development cycle, effectively reduces costs, and greatly improves development efficiency. Developers can accurately identify potential problems in the early stages of the project, so as to optimize the design and accelerate the iterative process.

Rapid prototype development in the field of automatic driving includes:

- Software prototype: Develop simulation environment and test automatic driving algorithm and decision logic.

- Hardware prototype: Build the physical components of an autonomous vehicle, such as sensors, actuators, and control systems.

- System integration: Integrate the software and hardware prototypes for real vehicle testing and verification.

The development and application of automatic driving are faced with the following challenges: diverse number/models of sensors, multiple communication protocols, high transmission rate requirements, high computing power requirements, large memory requirements and complex test environment, It also puts forward strict requirements for the platform environment of hardware and system construction:

· Adaptation and compatibility of various types of sensors

· Access capability of large bandwidth signal data

· High definition video processing capability

· High time synchronization accuracy

· Raw data access requirements

VCARSYSTEM's Autonomous Driving Data Recorder is the above challenge to provide systematic solutions

VCARSYSTEM's Autonomous Driving Data Recorder ADS 4000 Series — Software and hardware integration, high scalability, high-performance vehicle data recording&prototype development system

Product parameters：

·GPU : Support two pieces of RTX 30 Series NVIDIA ® Graphics card

·CPU : Support the 11th generation Intel Core ™ i9 LGA1200 CPU

·DDR : 4 SO-DIMM memory sockets, supporting up to 128 GB ECC/non ECC DDR4 2133 memory

·Camera : 12 way GMSL 1/2 camera 8M FPS HD camera access

·High synchronization accuracy : Up to 500 μ s-1ms synchronization accuracy, supporting PTP, GPS (PPS) synchronization

·High transmission bandwidth : Satisfy the storage of RAW/YUV and other original video data

·High processing performance : Rapid prototyping capability, support V4L2, Socket CAN, ROS and other data interfaces, and can directly access the original data of the sensor

·High adaptability : Support various types of sensor access in the adaptation market

Technical advantages

1. Open platform environment to adapt to dynamic requirements

ADS 4000 series is based on open Linux software environment, together with rich hardware interfaces and expandable boards, can synchronously access data related to mainstream sensors and vehicle ECUs, such as laser radar, millimeter wave radar, on-board camera, XCP/CCP, vehicle signal (CAN/CANFD/LIN/FlexRay/ETH), etc., to meet different customer needs.

2. High specification hardware platform, providing extreme performance

The ADS 4000 series, as a high-performance working platform, can achieve a synchronization accuracy of 500us - 1ms, support gptp, GPS (PPS) synchronization, meet the requirements of RAW/YUV and other original video data storage, and can build a rapid prototype of the domain controller, support multiple data interfaces, and directly access the original sensor data.

At the same time, the powerful GPU/FPGA hardware accelerator is used to develop, verify and optimize sensor fusion and perception algorithms and neural networks. Its scalability and flexible settings can customize the system according to requirements, and flexibly respond to changing requirements in the dynamic ADAS/AD development environment.

Application mode

ADS 4000 provides a powerful solution with high performance and scalability. Through direct connection and bypass application modes, it can meet the rigorous prototype development requirements of multi-sensor applications with a large number of environmental sensors, and enable it to record a large amount of data during driving tests, so as to playback, simulate or train neural networks in the later stage.