The introduction of Cruise Control, in 1948, was a significant step toward vehicle automation, but far from autonomous vehicle technology. In the early 1980s, Ernst Dickmanns and his team rigged a car with a set of cameras and 60 micro-processing modules to detect objects on the road, pioneering dynamic computer vision, a key technology of the driverless car.

The General Atomics Predator drone is one of the most successful unmanned planes of recent history. The technologies behind the Predator drone, such as radar that can see through clouds or smoke and thermal imaging cameras that allow travel by night, have been making their way to civilian autonomous vehicle manufacturers.

Autonomous vehicles cannot exist without high-performance computing. The technologies driving these autonomous vehicles have evolved through the participation of government agencies and private companies, including Boeing, Nasa, BMW, Google, and Tesla, just to mention a few.

Autonomous vehicles are powered by various technologies such as long-range radar, LIDAR, cameras, GPS, sensors, data storage systems, high-performance computers, and networks. Every single autonomous car continuously gathers, processes, and transmits enormous amounts of data, far too much for traditional technology and infrastructures to handle.

Driverless vehicle manufacturers recognize that the industry cannot move forward without implementing complex computing solutions. It will be impossible for autonomous vehicles to successfully detect road hazards and respond in a timely, appropriate manner without utilizing a technology capable of near-instantaneous analysis of huge amounts of data.

Related posts:

Deep Learning

Driverless cars were science fiction until researchers adopted an AI technique known as deep learning, which relies on GPUs (Graphical Processing Units). GPUs are a critical part of autonomous vehicle technology as they are engaged in the most complex calculations, like machine learning, AI, and deep neural networks. Deep learning is a must for autonomous vehicles because we can’t develop software that anticipates every possible scenario self-driving vehicles might encounter. GPU enables neural networks to learn from data. For example, to teach a deep neural network what a house looks like, we feed it thousands of house images, and it eventually “learns.”

Superhuman Sensing

High-assurance computing through HPC is a vital part of autonomous vehicle operations. Cameras and smart sensors capture a continuous stream of data from the vehicle’s surroundings, while HPC analyzes the data and uses it to ensure safety and efficiency. Object recognition and tracking, mapping and localization, movement prediction technology, and the AI algorithms and software that maneuver and control the vehicle all rely on HPC.

Fully autonomous vehicles require multiple redundant sensor systems: LIDAR, camera, and radar-based.

Light Detection and Ranging (LIDAR)

One of the most important technologies behind autonomous vehicles is LIDAR. LIDAR technology is not new. It was developed by NASA and the U.S. military in the early 1960s, and it has been deployed in industry and military operations ever since. LIDAR is critical for self-driving cars because it provides a 360-degree view of the surrounding area. A LIDAR system includes four major components: a laser pulse transmitter, a pulse receiver, an optical analyzing system for data processing, and a high-performance computer.

LIDAR emits thousands of laser pulses each second and measures how long they take to come back as they bounce off objects. The data from the LIDAR is processed by the onboard computer in a fraction of a second. There are two dominating trends in the automotive industry: infrared LIDAR systems use a rotating laser – leveraging MEMS (Micro-Electro-Mechanical System) – or a solid-state LIDAR.

From Cameras to Dynamic Vision

The world’s first truly self-driving wheeled vehicle dates back to the space race in the early 1960s. The vehicle was called, “The Cart,” and it was equipped with high-performance computers and cameras of the time. Cameras remain a critical component of the autonomous vehicles of today and the future. Autonomous vehicles require vision that detects, localizes, and classifies objects fast enough to feed that data to the rest of the driving system, so other units of the system can use the data to make real-time decisions.

Mono and stereo camera systems working together with radar technology provide an accurate assessment of speed and distance. They also outline obstacles and moving objects. There are radar sensors, short-range (24 GHz) or long-range (77GHz), at the front and back of the vehicle to monitor traffic.

Rear and 360° camera systems generally work within a centralized architecture. The central control unit processes the raw data of four to six cameras. Additional FPGAs (Field Programmable Gate Array) are required for hardware acceleration, requiring modern data compression methods and large storage capacities.


ADAS (Advanced Driver Assistance Systems) need several radar sensors that make a critical contribution to the overall operation of autonomous driving. Self-driving vehicles are using radars based on 24GHz or 77GHz; the 77GHz provides higher accuracy for distance and speed measurements. The advantages of 24GHz-based radars are smaller antenna size and lower interference.

Simulations of Petabytes of Data

Predictions indicate that, in just an hour and a half of drive time, autonomous vehicles will send over four terabytes of data to the cloud. This is the equivalent of around 1,000 DVDs. Current infrastructures cannot possibly manage the load of millions of vehicles generating terabytes of data on a daily basis. This level of data production, transmission, and analysis requires HPC.

Developers already use HPC to run thousands of test simulations involving petabytes of data on each piece of software that goes into a driverless vehicle. Much of this software provides the AI learning aspect that the vehicle relies on to read its surroundings and location, recognize and track other vehicles, objects, and pedestrians, and even learn how their passengers like to “drive.”

Faster Testing and More Sophisticated Analysis

Autonomous vehicle manufacturers are becoming increasingly reliant on HPC for complex testing, demanding simulations, and data analytics programs. As the number of smart sensors and high-resolution cameras on each vehicle grows, so does the size of the data sets that need to be transmitted, analyzed, and stored. Traditional IT infrastructures are incapable of handling the petabytes of data that need to be managed, and automakers are turning to HPC solutions that can provide the speed, reliability, and storage parameters necessary to meet data demands.

Data Management

Data storage demands will continue to grow as self-driving car manufacturers increase development and production. These companies must implement systems that are capable of scaling up to accommodate the petabytes of data that will be generated by advanced cameras, sensors, and software. Autonomous vehicle developers need data management systems that offer a single point of access and allow for fast retrieval of enormous amounts of data.

Machine Learning and AI

Driverless cars utilize advanced AI technology to analyze their surroundings and immediately determine the best course of action to ensure efficiency and passenger safety. The ability to recognize and respond appropriately to road signs, other vehicles, pedestrians, and potential hazards requires extremely advanced sensors, high-level processing power, and the capacity to learn and adapt.

Roads are unpredictable places, and pre-programmed responses cannot accommodate for the unexpected. Autonomous vehicles must be able to continuously learn and make optimal decisions under all circumstances. HPC powered AI can meet these deep-learning needs at a reasonable price point.

Biometrics and Recognition Technology

Despite incredible technological advancements, there are challenges that remain in autonomous vehicle development. Until all of these issues are solved, the human passenger needs to stay alert and be prepared to manually take over in critical situations where the vehicle is unable to perform correctly. Self-driving vehicles must be fitted with monitoring systems that include biometrics and recognition technology to ensure that the occupant is awake, medically sound, and capable of operating the vehicle if the need arises. If these conditions are not met, the vehicle will safely maneuver and park itself.

Tiers of Data Infrastructure

Autonomous vehicles will require access to multiple infrastructures to manage the terabytes of data that each vehicle generates. This will include:

  • Onboard computers – connecting to supercomputers that can process all of the data coming from the vehicle’s cameras and sensors
  • Wireless connectivity – to communicate with towers, other vehicles, and smart infrastructure elements
  • Edge data centers – for data analytics
  • Core data centers – providing in-depth data evaluation

HPC is a cost-effective solution to manage the massive jump in data production that will be part of the ecosystem with the widespread use of driverless vehicles. In addition to the convenience factor, autonomous vehicle development is also poised to begin rapidly decreasing the high number of auto accident-related deaths occurring every day around the world. However, the safety measures and optimal effectiveness of self-driving vehicles is unattainable without the use of HPC solutions. HPC is capable of handling the petabytes of data generated by autonomous vehicles and powering the advanced AI programs that will ensure reliability, smart decision-making, and passenger safety.