How AI/ML applications are transforming the automotive value chain
AI driven ADAS (Advanced Driver Assistance Systems) technologies are definitely going to be a ‘standard feature’ in each vehicle that would come out from production line very soon.
Artificial Intelligence and Machine Learning (AI/ML) applications are revolutionising the automotive sector. From concept to engineering to testing to manufacturing to sales and after sales, we find its influence and impact everywhere.
Five to six years ago, we used to find AI/ML applications mostly in web and e-retailer domain from the tech powerhouses like Google or Amazon. Then slowly people started thinking about the of possibilities of taking it to the engineering and manufacturing domain and leveraging the power of AI. This became a reality when businesses got a sense of the power of data insights and decided to capture critical information in a structured and disciplined manner. Today we find most of the research and innovation projects are powered by AI/ML, not just because the complex algorithm libraries became the open-source but also because of sheer computation ability of GPU/TPU (graphical processing unit and tensor processing unit) powered infrastructure quickly churning huge amount of data, giving multiple possibilities of configuration and optimisation.
Autonomous Vehicles and ADAS
Today, car manufacturers are racing to develop the technology to make fully self-driving cars a reality. Advances in processors, camera technology and AI have brought us closer than ever before. Developers of self-driving cars use vast amounts of data from image recognition systems, along with machine learning and neural networks, to build systems that can drive autonomously. Though the ‘full driving automation’ where the driver does not exist anymore and the human being becomes a passenger, may not be realistic or very popular in countries with huge traffic congestion, but the AI driven ADAS (Advanced Driver Assistance Systems) technologies are definitely going to be a ‘standard feature’ in each vehicle that would come out from production line very soon.
A lot of work has already been done in automotive safety, starting from automatic emergency braking, pedestrian detection, surround view, parking assist, lane departure warning, and so on, and it seems that the sky is the limit. AI algorithms can analyse sensor data to identify potential dangers in real-time, which mitigates the risk of accidents. AI-based safety features, such as blind-spot detection and adaptive cruise control, use sensors and cameras to detect vehicles in the driver's blind spot and monitor the distance between the car and the vehicle in front. Various safety features starting from “driver distraction/drowsiness/rash driving alerts” to “advance road irregularities notification” (potholes, speed breaker, etc.) are also part of AI-powered safety systems.
Driver and passenger comfort through personalised experience is also an emerging area of application of AI where successful use cases range from interactive voice command to configuring a seat relaxation function, altering the interior lighting, music, air-conditioning, illumination intensity etc.
Product Design and Development
Having talked about product (vehicle) specific and customer (driver, passenger) experience centric applications, AI is also revolutionising the way product design, development and production are carried out. The Auto OEMs are extremely concerned about the cost optimisation and productivity improvement to stay ahead in the market and hence, using AI extensively to improve underlying processes, saving on material resources and improving on time to market.
And no regret, AI has not disappointed them. It is going to be a reality soon when Generative Adversarial Networks (GANs) will start throwing multiple vehicle concept sketches basis a few input requirements for quick evaluation. Various cognitive models are already in action to bring down NVH (Noise, Vibration & Harshness) in the vehicle, thus improving overall vehicle quality. A driver assistance system scales Real Driving Emission (RDE) trip validity from existing 35% to 80% and thereby reducing RDE testing expenses by 44%. Predictive failure or underperformance of components has led to consumers to stay alert and helped in informed decision making to reduce overall turnaround time, particularly in commercial vehicle space! For example, Tyre wear and engine nozzle failure issues during the warranty period are reduced by 20% and 35% respectively, thanks to real-time field data made available through ‘Connected Vehicle Platform’.
AI is redefining manufacturing practices with implementation of Industry 4.0 that are otherwise impossible to achieve. IoT and AI are truly revolutionary technologies that use real-world data to improve business practices and help companies make better decisions. Below are the ways that AI help automotive companies reduce costs and optimize manufacturing processes:
a. Quality Control
b. Predictive maintenance
c. Production optimization and integration
d. Digital twin
AI analyses historical operational data of an equipment and compares with its current state of operation to forecast whether the equipment performance is optimal or needs maintenance. It can even establish relationship between equipment operational data and failure instances to predict Remaining Useful Life (RUL) of the equipment or its components. Avoiding unplanned downtime, reducing maintenance cost and improved asset utilisation are the key business benefits here.
Total Cost of Ownership
TCO has always been a concern for the fleet owners and fuel economy remains a universal expectation. To address that several models are being worked upon incorporating manufacturing parameters, powertrain calibration as well as consultancy on driving patterns and behaviours.
Sales and Service: AI is being imbibed into sales and marketing, giving a boost to retail prognosis, customer propensity, and likelihood of deal conversion. Service and warranty are the two major areas where there are lots of scope for AI driven decision-making, automation in claim validation and brining out actionable insights for all stakeholders like suppliers and engineering and plant quality teams.
NextGen Generative AI and Foundation Models
Traditionally, people used to build ML models using historical labelled data and used it for prediction purposes. Now with the introduction of Generative AI, the new trend has become working on huge unlabelled data on a self-learning mode and let thus created Foundation Models adapt to prediction and other decision-making tasks. Clearly, there is a shift from traditional ML models to Foundation models.
Advancements in AI, such as reinforcement learning and deep neural networks, will continue to shape the evolution of autonomous vehicles and connected cars. Further integration of AI into the automotive ecosystem will result in enhanced safety, improved energy efficiency, and a seamless driving experience.
Avijit Santra is a Program Manager at Tata Technologies. Views expressed are those of the author.
The experience of owning electric bikes is new, and remaining informed, aware and proactive will help in becoming better...
By ensuring that only qualified drivers occupy the roads, authorities can substantially reduce the risk of accidents ste...