ADAS Data Capture & Storage Computing Solutions

ADAS Data Capture and Storage Computers (Advanced Driver Assistance Systems) 

It should come as no surprise to those familiar with autonomous vehicles and vehicles equipped with advanced driver assistance systems (ADAS) that such vehicles collect and need a large amount of data to train deep learning and machine learning algorithms that assist cars with driving, avoiding obstacles, and avoiding accidents. Training such systems typically requires a system capable of capturing and storing real-world data to train the advanced driver assistance system at a later time. 

Capturing data generated from cameras and sensors is not an easy task and requires powerful computer systems that are equipped with powerful processors and plenty of high-speed data storage. The more real-world data the manufacturers of advanced driver assistance systems have, the more accurate the system will detect objects and drive the vehicle. 

Typically, vehicles with advanced driver assistance systems, such as automatic braking, collision protection, lane keep assist, adaptive cruise control, and autopilot technology, are equipped with one or a combination of the following items: high-resolution cameras, Lidar, ultrasonic sensors, sonar, GPU, and other types of sensors that help the vehicle “see” the surrounding environment. To train algorithms, test vehicles must be equipped with ADAS data capture and storage computers that are capable of connecting to and capturing information from vehicle cameras and sensors. 

For a solution to capture and store data, it must be equipped with robust processing power and plenty of high-speed storage. This is so because vehicle cameras and sensors generate Terabytes of data that’s needed to train the algorithm. Some estimates state that vehicles equipped with cameras and sensors typically generate 4TB to 5TB of data per vehicle per day. So, having a system with enough storage to store the data is crucial for the system to keep up with the voluminous amount of data generated by cameras and sensors. 

Furthermore, the system that stores the data must be onboard the vehicles because it is challenging, if not outright impossible, to collect and send all of the data generated by the sensors and cameras to the cloud. So, to capture and store the data, organizations will require an edge computing solution that has the required performance and storage capacity.

Premio offers a variety of AI edge inference computers that are specifically designed for deployment in vehicles to collect and store camera and sensor information. For example, Premio’s ADAS data capture computers can be configured with the powerful 8
th and 9th generation Intel Core i3, i5, and i7 Processors, offering organizations plenty of performance for collecting and storing high-resolution camera and sensor data. Furthermore, ADAS Data storage computing solutions can be configured with plenty of storage for storing camera and sensor data. 

For example, ADAS data capture and storage computers can be configured with multiple M.2 NVMe SSDs, U.2 NVMe SSDs, regular SATA SSDs, and HDDs. Some models offer support up to 8x U.2  NVMe SSDs or  2x internal SSDs or HDDs, and 2x hot-swappable SSDs or HDDs in the SATA protocol. The flexibility and amount of storage that can be added make Premio’s vehicle data capture and storage computers more than capable of storing the Terabytes of sensor data generated by autonomous vehicles and vehicles equipped with advanced driver assistance systems (ADAS). 

How Much Data Do ADAS Equipped Vehicles Generate?

Autonomous vehicles and vehicles equipped with ADAS can generate anywhere from 4TB to 5TB of data per day. This is so because vehicles with ADAS are often equipped with several cameras and sensors that include sonar, ultrasonic, LiDAR, and GPS sensors. The cameras on vehicles equipped with ADAS typically generated 20-60 MB/s, radar generates 10KB/s, LiDAR generates 10-70 MB/s, and GPS generates approximately 50KB/s. When you combine all of these numbers, your vehicle is generating up to 130MB per second of data which adds up to approximately 8GB of data per minute.

Image Source: Synopsys

So, autonomous vehicles that are on the road for 10 hours a day can generate 4.8TB of data per day. The more sensors and cameras a vehicle is equipped with, the larger the amount of data that’s generated. Of course, these numbers can vary based on the resolution of the cameras that are used and the number of sensors and cameras that are being used by the system. Some systems may use less, while others may have more. 

What Happens to ADAS Data That is Captured and Stored? 

The data that is captured and stored by advanced driver assistance systems (ADAS) is used to develop and improve the performance of ADAS systems. For example, if an ADAS uses deep learning or machine learning, the data that’s collected by AI edge inference computer is used to train the ML (machine learning) or DL (deep learning) model so that its better able to detect objects, persons, lane markers, and streets signs. Overall, the more data that’s used to train the model, the better the model will perform when exposed to environments and objects it has never seen before. 

Furthermore, during the training phase, artificial neural networks are taught to classify certain objects or properties in the same manner that humans classify them. Training a model requires a ton of processing power; therefore, it’s usually performed in data centers with the help of GPUs. GPUs can accelerate training because they can process significantly more data than CPUs. This is so because GPUs have significantly more cores than CPUs, allowing them to process significantly more data simultaneously. After a model is trained, it’s usually deployed in a vehicle to perform inference analysis on new data (environments and objects) that it has never seen before. The better the training, the better the algorithm will be at identifying objects and driving the vehicle.

Deep Learning Training - Image Source: Intel

That said, to capture and store the large amounts of data used to train models, test vehicles must be equipped with hardened AI edge computers that are capable of connecting to and capturing information from the vehicle cameras and sensors. AI edge inference PCs come equipped with powerful processors and robust storage solutions that are capable of storing the Terabytes of data generated by vehicle sensors and cameras. 

Powerful edge computers are required because data must often be stored and processed locally, only sending critical processed information to the cloud. This is so because sending Terabytes of data via a cellular connection is both very difficult and expensive. It’s difficult because cellular carriers do not provide sufficient upload speeds to upload the large amount of data generated by vehicle sensors.

Also, the second obstacle to sending all raw data to the cloud is that sending so much data via a cellular connection is super expensive, making it cost-prohibitive for most organizations. So, AI edge inference computers are used to store the data, offloading some processed data to the cloud via cellular connectivity while offloading the bulk of the data at a centralized location by physically removing hard drives and manually offloading the data onto a central computer. 

Advanced Driver Assistance Systems (ADAS) Ruggedization & Features 

AI edge inference computing solutions are hardened to withstand deployment in challenging conditions that are too harsh for regular desktop computers. We will discuss how edge AI computers are hardened to withstand deployment in volatile environments. 

1. Fanless Design 

When searching for an ADAS computer that can capture and store the data generated by vehicle cameras and sensors, you should select a fanless solution. You should select a fanless solution because fanless industrial computers are best suited for deployment in vehicles where they will be exposed to dust, debris, and other small particles. The fanless design keeps dust and small particles from entering the system and damaging components.

Furthermore, the fanless design eliminates the use of fans from the system. The elimination of fans creates a significantly more reliable system because fans are the leading cause of many electronics malfunctioning, including computers. So, by eliminating them, we have eliminated a common point of failure, making the solution much more reliably and durable.  

2. Shock and Vibration Resistance 

Furthermore, when searching for an ADAS PC to capture and store vehicle sensor data, you should select a solution that is equipped with shock and vibration resistance. AI edge inference computers come with 50Gs of shock protection and 5GRMs of vibration resistance in compliance with the MIL-STD-810G. Shock and vibration resistance allows systems to be deployed in vehicles where they will be exposed to frequent shock and vibration as cars travel on roads.

Additionally, Premio ADAS computers are made shock and vibration resistant by eliminating all cables from the system. The removal of all cables from the system reduces the number of moving parts, thus reducing the number of parts that can fail, creating a more reliable solution.

3. Wide Operating Temperature Range 

Vehicles are mobile, and so they might travel to environments that are extremely cold or extremely hot. As such, when selecting an ADAS data acquisition system, you should choose an option that can handle exposure to extreme temperatures. Premio’s AI edge inference computers have been designed and built to withstand exposure to extreme temperatures. In fact, they have a wide operating temperature range, ranging from -25⁰C to 60⁰C, making them ideal for mobile deployments in vehicles. 

Premio AI edge inference PCs have a wide operating temperature range because they are fanless, and they are configured using wide temperature range components that are specifically designed to survive challenging deployments in environments that experience fluctuating and extreme temperatures.  

So, whether the vehicle travels to the Mojave Desert where the temperature reaches 50⁰C or in New York during the winter where the temperature reaches -15⁰C, our AI edge inference solutions will operate optimally and reliably even when exposed to such extreme temperatures. Also, it’s worth noting that you will not need to invest in additional hardware to achieve the wide operating temperature as the system has a wide operating temperature range out of the box. 

As such, rugged edge computers are capable of handling exposure to extreme temperature, something that regular desktop computers are not capable of handling. This is so because regular, consumer-grade computers are not made from wide-temperature components, and they aren’t designed to handle exposure to extreme temperature. They are made for home or office use in temperature-controlled environments and not challenging in-vehicle deployments, as are rugged edge computers.  

Typically, desktop computers have an operating temperature that ranges from 5⁰C to 40⁰C, whereas rugged edge PCs have a wide operating temperature, ranging from -25⁰C to 60⁰C, making them significantly more capable of performing reliably and optimally in challenging environments that experience extreme temperatures. 

4. Power Input Compatibility 

When searching for an ADAS computing platform, you should select an option that is equipped with a wide power range. This is so because ADAS computers deployed in vehicles must be able to run from a vehicle’s power. For example, Premio edge computing solutions have a wide power range, allowing the system to be powered from a variety of different power input scenarios. Additionally, Premio’s solutions come equipped with a variety of different power protection features that include overvoltage protection, surge protection, and reverse polarity protection.  

5. Power Ignition Management 

Furthermore, when selecting an ADAS data acquisition system, you should select a solution that is equipped with power ignition management. Premio’s AI edge inference computing solutions come equipped with power ignition management capabilities, enabling the system to detect when a vehicle has been powered on, sending a signal to the system to begin a boot delay. Also, the system detects when a vehicle has been turned off, allowing it to perform a delayed turn-off. Delaying the shut down of the PC allows the system to finish working on its current task, preventing data loss or corruption. Moreover, when the system is turned off, the power ignition management features prevents the edge computing solution from draining the vehicle’s power. 

6. CANBus Capable 

CANBus is a protocol that delivers messages between the various components of a vehicle. Tapping into the CANBus system allows organizations to gather information that includes vehicle speed, RPM for the engine, throttle position, steering angle, tire pressure levels, and a variety of other critical vehicle information. So, when selecting a solution, you should look for a system that’s capable of tapping into the CANBus network to collect and store vehicle information. For example, Premio’s AI edge computers can tap into a vehicle’s CANBus system, collecting data from the various sensors and devices connected to the CANBus network. The information collected is invaluable because it can be used to develop advanced driver assistance systems. 

Although CANBus can be found in almost all vehicles currently on the road, some organizations are exploring the option of equipping their vehicles with automotive ethernet. Automotive ethernet (100 Base-T1) will be adopted in the future due to the increase in bandwidth that’s required in connected cars, autonomous vehicles, and self-driving vehicles.

7. Wired & Wireless Connectivity 

Edge AI inference computers can be configured with a variety of wired and wireless connectivity options, allowing your system to connect to the internet and other devices that utilize wired and wireless connectivity. For example, all AI edge inference computers are equipped with Dual RJ45 Gigabit ports, offering organizations the ability to connect transfer data at extremely high speeds and connect to high-resolution cameras and sensors.  

Moreover, ADAS edge computer systems are equipped with wireless connectivity via Wi-Fi 5 or the latest generation Wi-Fi 6 modules. Wi-Fi is excellent because it offers organizations plenty of flexibility when determining wireless connectivity speed and range. Also, two Wi-Fi 6 technologies enable devices to connect to numerous IoT devices, and these two technologies are Mu-MIMO and OFDMA.  

Mu-MIMO, short for multi-user multiple-input technologies, allows edge computers to connect to multiple Wi-Fi-enabled devices simultaneously. Mu-MIMO has the ability to significantly increase a network’s throughput, making it great for high-density networks. Mu-MIMO is available on both Wi-Fi 5 and Wi-Fi 6. Mu-MIMO is an improvement compared to SU-MIMO, which enabled single-user MIMO, allowing a device to send/receive data to a single device simultaneously. Mu-MIMO expands the technology to multiple users. 

OFDMA, short for orthogonal frequency division multiple access, divides Wi-Fi channels into smaller frequency allocations known as resource units, allowing your device to communicate with multiple clients simultaneously. 

That said, since vehicles are often moving, it’s difficult to connect them via Wi-Fi and wired connectivity options, necessitating that devices have cellular connectivity to offload critical information to the cloud. AI edge inference computers come equipped with Dual SIM sockets, allowing two cellular data carriers to be added to a device for redundancy. If one of the data carriers is unavailable in a remote environment or the signal is poor, the system can be programmed to connect to a secondary cellular carrier to offload critical data to the cloud. 

Also, AI edge computers can be configured with Bluetooth connectivity. Bluetooth offers reliable one-to-one connectivity and many-to-many connectivity. That said, Bluetooth does not have the range and speed that Wi-Fi and wired connectivity offer. Nevertheless, they do provide simple and reliable connectivity to sensors and IoT devices. 

Bottom Line 

The bottom line is that vehicles equipped with ADAS require a lot of data to train and develop advanced driver assistance systems that can safely operate a vehicle and prevent accidents. To capture and store data, vehicles must be equipped with AI edge computers equipped with powerful multi-core processors equipped with large amounts of storage to store the Terabytes of data generated every day by each vehicle. Furthermore, ADAS computers are equipped with a rich I/O enabling systems to connect to many sensors and cameras to collect and store sensor data. If you have any questions or need assistance selecting an ADAS data capture and storage PC, please contact one of our edge computing professionals, and they will be more than happy to assist you with selecting a solution that meets your specific requirements.