By Dr. Jose Pereira, Director at Frost & Sullivan
The automotive industry is undergoing a period of mass transformation that has never been seen before, with many challenges to face. It is transforming its development processes to turn historically hardware-centric vehicles into fully connected products that can be updated even after they leave production lines.
Moreover, climate change concerns are driving engineering complexity due to extensive air quality-related vehicle CO2 regulations and evolving consumer demands. New entrants and start-ups have also created a new enthusiasm for autonomous driving technologies and other novel in-vehicle cockpit experiences.
More recently, the COVID-19 pandemic unlocked another challenge with the semiconductor supply chain crisis. Additionally, the related lockdowns wreaked havoc on the industry’s production planning and global vehicle deliveries. These combined technology challenges and customer expectations create huge financial burdens for leading players as they try to keep up.
Due to this challenging context, manufacturers are already taking drastic actions to free up the resources needed to navigate these difficulties, such as cutting back their portfolio’s range of vehicle models. The goal is to reduce fixed costs by improving operational efficiency and being selective about products and technologies.
Leaders should pay attention to the potentially transformative role of machine learning in making engineering work more efficient. Tier 1 suppliers and original equipment manufacturers (OEMs) will need to reorganize their research and development teams and supplier relationships to develop these new connected and intelligent vehicles more efficiently. Enabling engineers to develop products faster and with fewer resources has become a key priority.
Building the road to artificial intelligence
Leading manufacturers are taking different approaches to the ongoing transition, but there is already an increased awareness and interest in the field of artificial intelligence (AI), leveraging the power of machine learning technology. For example, VW launched a dedicated electric vehicle family and is investing heavily in building up its software development capabilities to explore use cases for machine learning. This will improve software development and vehicle functionality.
Toyota is betting on a continued range of powertrain technologies, including hydrogen and fully electric vehicles, that will use machine learning to optimize performance and emissions. It is developing its AI capabilities via a new organization named Woven Planet and has acquired various companies to build its expertise in operating systems, mapping solutions, and self-driving.
Hyundai is not lagging. It’s now offering super-smooth electric vehicles to match South Korea’s digital ambition, and it’s already using machine learning in various applications, including providing optimum gear shift predictions using real-time vehicle and environment data. In addition, major suppliers like Bosch, Magna, and Denso are also evolving their competencies and capabilities.
Earlier this year, VW’s software unit, Cariad, and Bosch announced that they would work together on Level 2-3 assisted driving technology to catch up with their rivals. Denso has also collaborated with major organizations like Toshiba to develop a machine learning system for image recognition to help achieve advanced driver assistance and automated driving. It is also exploring other applications via partnerships with organizations like Canada’s Institute for Data Valorization (IVADO).
Finally, Magna has invested in AI R&D institutes like Vector in Canada and AI companies like Seeing Machines to support vehicles and manufacturing-focused innovations.
Many of these innovations are enabled by the massive increase in cloud processing power and rapidly improving machine learning algorithms. However, most applications rely predominantly on approaches like deep neural networks (DNNs), which require huge datasets to train models and achieve robust predictive performance for tasks like assisted driving. The result has been an exponential rise in computing power requirements and the cost to train models that are creating new inefficiencies. As a result, innovators are addressing these challenges with new approaches to machine learning that are more efficient and effective, requiring much fewer data to produce reliable models of complex system behavior.
New advanced machine learning techniques
For AI to have a much greater impact on the automotive industry in many areas beyond autonomous driving, it must become more accessible. That means a radical reduction in data requirements for training models to deliver robust performance and much easier interaction with tools and techniques so non-artificial intelligence experts can also benefit from these technologies and apply them to their unique challenges.
Such technologies could have a dramatic impact on the efficiency of current product development processes. While the industry has used computer-aided design and engineering for years to optimize designs and simulate system performance, as these systems increase in complexity, they are becoming more challenging to optimize with traditional physics-based tools. New machine learning methods offer the opportunity to give engineers additional powers to tackle these complex problems faster.
To address these shortcomings, Secondmind developed the Active Learning platform. The company is already unlocking the potential of machine learning for Japanese OEM Mazda and other companies in the automotive design and development value chain.
The Secondmind Active Learning platform uses mathematical simulation to help engineers optimize complex system designs at an early stage and requires 80% fewer data to create high-precision models. By utilizing a platform such as this, OEMs can significantly reduce the duration of the design and development life cycle by eliminating many design iterations and time spent simulating. Recent applications in powertrain testing have demonstrated a more than 50% reduction in testing times required to achieve robust calibrations, effectively doubling the efficiency of that expensive development process step.
OEMs should work closely with machine learning specialists like Secondmind to enable efficiency gains that directly impact the return on investment for small design teams. Furthermore, these machine learning solutions will assist OEMs in achieving their business goals while empowering their engineering teams with cutting-edge technologies that improve innovation and reduce costs with minimal training.
To learn more about how cloud-based machine learning can accelerate complex automotive processes, please access the Frost & Sullivan Executive brief here.
