singapore skyline

Navigating the Life Cycle Challenges of AI in Vehicle Systems

Enabling revolutionary progress

Enabling revolutionary progress

NAVIGATING THE LIFE CYCLE CHALLENGES OF AI IN VEHICLE SYSTEMS

Artificial Intelligence (AI) is a transformative technology that is reshaping organisations and redefining competitiveness in this increasingly complex business landscape. This holds relevance in the automotive industry where the integration of sophisticated AI-powered functionalities into vehicle systems becomes more prevalent. Each addition of an AI-based system into the vehicle adds multiple layers of complexity. With increasing complexity, it becomes very difficult to ensure the safety and security of the vehicle as a system.
 
This article highlights some of the challenges the automotive industry faces when dealing with AI system development and covers possible solutions.

 

Integrating AI Lifecycle with the Automotive Development Lifecycle

One of the primary obstacles confronting the industry revolves around effectively integrating the AI lifecycle into the vehicle development lifecycle. While the automotive sector adheres to the well-established V-model product development lifecycle, this approach is not as straightforward in the context of AI.

The AI development lifecycle is distinct and comprises two additional key phases: the data lifecycle and the model development lifecycle. Successfully integrating these into an organisation's standard automotive development lifecycle will be crucial for safe and robust AI-powered automotive technologies.

To integrate these lifecycles effectively, coordination and synchronisation in both the AI considerations and development are essential. For instance, during concept development, identify and define AI use cases and requirements. In the design phase, develop and integrate AI algorithms and models into the vehicle's software and hardware architecture. Testing and validation should include procedures to assess AI system performance, robustness, and safety. The integration should also consider continuous monitoring and maintenance of the AI system throughout the vehicle's lifecycle. This involves collecting and analysing real-world data to improve AI models and ensure optimal performance in various driving conditions.

 

Defining AI requirements based on system requirements; valuing traceability

We have observed that the requirements for an AI system are often incomplete or inconsistent with those expected of the vehicle’s systems. This disparity becomes particularly apparent when integrating an intricate automated driving system into the vehicle.

Defining AI requirements based on system requirements ensures that the AI effectively contributes to the desired outcomes of the overall system. This enables the development of AI solutions that are tailored to the specific needs of the automotive industry, be it enhancing safety, improving efficiency, or enabling advanced driver assistance features. Traceability – also crucial in this definition - refers to the ability to track and document the relationships between different requirements, design decisions, and system components. In the automotive industry, traceability ensures that there is a clear understanding of how the AI requirements are derived from the system requirements. It assures that every AI requirement is well-mapped and connected to its corresponding system requirements, enabling transparency and accountability throughout the development lifecycle.

 

Ensuring Automotive Safety: Performing AI Risk Assessments

Incorporating AI technology into vehicles brings forth new complexities and potential risks. These risks primarily revolve around data and model-related factors. Data risk concerns the quality, quantity, and relevance of the data utilised, as it directly influences the performance and safety of AI systems in vehicles. On the other hand, model risk encompasses the uncertainties and limitations associated with the AI models used in automotive systems. Even when high-quality data is available, there are inherent risks involved in the model development process.

self driving AI

Performing AI risk assessments highlights the proactive nature of the approach. It involves conducting systematic evaluations to identify possible hazards and associated risks that AI systems may introduce. These assessments consider a range of factors, such as the performance, reliability, and robustness of AI components, as well as potential interactions with other vehicle components.

 

Treating AI Components as Configuration Items

The challenge with AI components (data, algorithm, and models) as configuration items (CIs) lies in their dynamic and ever-changing nature. Unlike traditional software components that are relatively static, AI components such as datasets, neural networks, or machine learning models are continuously evolving and are adapted based on new data and retraining.

employees working

Treating AI components as individual configuration-management systems will create an oversight in the lifecycle of CIs through the processes and tools adopted. Proper configuration involves selecting and integrating the most suitable algorithms and models, training them on appropriate and representative datasets, fine-tuning their parameters to achieve optimal performance, and finally establishing version control. By recognising the importance of AI components as CIs, the automotive industry can ensure the effective and efficient deployment of AI systems in various applications. 

 

Nestling AI Components Under the Change Management System

The evolving nature of AI components demands a meticulous approach to tracking system changes and updates. These components, being at the forefront of cutting-edge technology, constantly mature to optimise performance, efficiency, and safety. Within the organisation, staying informed about these modifications is of utmost importance.

By bringing AI components under the change management system, the automotive industry aims to implement structured processes for handling changes related to AI technologies. It helps maintain control and transparency over the changes introduced to AI systems, ensuring that modifications are properly reviewed, approved, and documented. This is particularly crucial for safety-critical systems in vehicles. It also ensures that changes to AI components are properly tracked, documented, and associated with the corresponding requirements, design decisions, and system configurations. This helps in understanding the rationale behind changes and enables effective communication among stakeholders.

 

Conclusion

As the automotive industry embraces the transformative power of AI in vehicle systems, it encounters a host of challenges that demand careful navigation and innovative solutions. The integration of AI lifecycles with the established vehicle development lifecycle poses a primary obstacle. 

.AI in automotive

Following this, defining AI requirements based on a vehicle’s system requirements is necessary for maintaining transparency and accountability for the developer or manufacturer. Safety also remains paramount, necessitating AI risk assessments to proactively identify and mitigate potential hazards associated with data and model-related factors. By addressing these challenges and implementing effective solutions, the automotive industry can confidently harness the benefits of AI while ensuring the safety and security of AI-powered vehicle systems.

Related resources

Artificial Intelligence

Establish trust, stay relevant, and utilise the full potential of AI

Learn More

Artificial Intelligence Quality Training

A practical guide to gain the required knowledge to master AI quality

Learn More

AI Quality Readiness Analysis

Check your organisation's AI quality readiness

Learn More

AI Quality Guided Assessment

A Systematic approach to assure AI quality and compliance

Learn More

Next Steps

Site Selector