Autonomous Mobility

Artificial Intelligence in Autonomous Vehicle Development and Testing

By Swapnil Nanir - Dec 08 2020
Autonomous vehicles need not only to understand their passengers and drivers but to be able to navigate a complex world. Artificial Intelligence or AI plays crucial role in perception, analyzing, annotating, and in decision making for an autonomous vehicle.

This is a mission-critical use case for AI, where there’s very little room for error. The stakes are very high as it concerns with human lives.

Vehicles with conditional automation (L3) and above need a much-sophisticated decision-making process. For instance, in a conditional automated car, the system has the make decision related to automated braking, steering control, and acceleration control with some degree of human intervention (only in certain cases). To enable this, the sensor suite of the vehicle senses the vehicle environment in 360 degree using cameras, radars, and LiDARs. However, AI currently is a child which needs to understand the world using these vision sensors. Understanding comes from learning what the environment contains - be it another vehicle, pedestrian, road markings, traffic signs, cyclist, localizing the vehicle, and so on.

To enable AI processor both hardware and software on-broad, perception systems of camera, LiDAR, and radar; localization and mapping systems, and V2X hardware and software system are required in conjunction. 

From the market perspective, the industry is witnessing surge in the use of AI and machine learning algorithms for higher level of automation in the vehicles. The graphs below give the clear trend on how the market for ADAS is increasing and how the penetration of AI hardware and software in the development of ADAS and autonomous vehicles in gaining huge momentum. 

M14 Intelligence _ Artificial Intelligence in ADAS and Autonomous Vehicle Development

M14 Intelligence - Artificial Intelligence Market in ADAS and Autonomous Vehicle Industry

The use of AI in automotive industry is increasing exponentially, where the traditional automotive players are implementing AI in research and development processes, supply-chain, manufacturing, marketing & sales, customer/driver experience & safety, connected & shared mobility services, and information storage and processing.

For ADAS and autonomous vehicles, AI is majorly implemented in two ways - one is in enhancing the in-car experiences and the other is in providing best-class safety.

In-cabin experiences using AI are currently in the form of driver and passenger monitoring systems, facial & gesture recognitions systems, and voice-enabled virtual assistants driven by ML and NLP.

The other important implementation is in the development and testing of autonomous vehicle systems. AI is one of the crucial elements for level 3 and above autonomy. While some companies see level 3 functionalities as an evolution of classic ADAS functions, which can be mastered with rule-based coding, level 4 and 5 autonomy require artificial intelligence to cope with the complexity of traffic situations. The latter typically demand large data sets (e.g. raw sensor data) for training, testing and validation of (deep learning) algorithms. In order to store and process these data, companies make use of data centers or cloud solutions. The data are labelled, clustered and ultimately used to optimize and update algorithms. It remains an open challenge today to efficiently validate artificial intelligence algorithms such as CNNs. AI algorithms operate like a black box in the sense that it is not trivial to determine what triggers certain decisions. Validating correct functionality is cumbersome and today only feasible statistically via numerous test cases.

Besides, data centers constitute the basis for simulation purposes. Theoretical estimates show that in order for level 3+ autonomous vehicles to achieve approval for commercial use, the system needs to undergo billions of kilometres of testing. It is neither economical nor does it prove to be a swift approach to achieve this amount of mileage in real world testing. Simulations effectively contribute to this requirement, covering more than 95% of the mileage demand. However, it remains a challenge to set up the proper test concept and collect or create sufficient data for validation. Overall, the challenges for companies working on autonomous driving technology are significant and vastly affect the dynamics of the automotive industry. 

machine learning
artificial intelligence
autonomous vehicle

Share this article

Related Posts


Competition in AI-powered AMR market

Competition between the leading players in the market or OEMs One of the main strategies implemented by major robotics manufacturers is acquisiti...

By Faith Wambui - Feb 03 2023


Funding in Autonomous Vehicle Technology- 2021 Outlook

Self-driving vehicles have been the most challenging technological innovation of this generation. Today, a large number of cars sold come with driver...

By Eva Sharma - Feb 02 2022


Skyrocketing Demand - Active Driver to Occupant Monitoring System Market

The push for safety by the regulatory bodies and OEMs across Europe, Asia, and United States by mandating driver monitoring and left-child detection,...

By Swapnil Nanir - Mar 08 2021


ADAS and Autonomous Vehicles Software, Electrical and Electronic Components (E/E) Market

ADAS (especially the level 2 and level 2+ solutions), autonomous driving (level 3 and above private vehicles and urban slow-moving robotic vehicles ...

By Swapnil Nanir - Jun 16 2021