AI: a means to an end

Connected and automated driving (CAD), I believe, is a means to an end. Research and innovation on CAD, pilots, living labs and - at the end of the day, of course - deployment of CAD technologies, should all contribute to high-level and overarching targets, like those set in the EC Sustainable and Smart Mobility Strategy (2020) or, more recently, Fit for 55.

Any new technology introduced should contribute to making our mobility system safer, more sustainable, resilient and inclusive. Key to achieving this will be the user-centricity of our solutions, and a systems approach. This drive for a systems approach is an essential part for the European CAD activities, and an inherent part of their success. Within the overall portfolio of technologies for CAD, artificial intelligence (AI) can be a very powerful tool to achieve several of these targets, as well as to combine ambitions. CAD should not be safe or resilient, but both. It should be sustainable and inclusive. In any combination. AI can tie the knots to achieve this.

AI can be excellently used in complex situations; there it is at its best. Our (road) mobility system is full of such complex situations. Many of these we experience in our daily trips. But when trying to achieve the overarching ambitions, we do need to take into account several more layers of information and aims. Decreasing congestion by optimised traffic management, greening of road mobility by more efficient use of transport, enabling seamless multimodal trips, enhancing road safety by using information of what is yet beyond our direct vicinity; they all add to the need of a systems perspective.

Real-time data

Using real-time data and shared data is essential. How this can be done has been (and will be!) the topic of research and discussion for quite some time. The CAD Knowledge Base supports this, bringing together the lessons learned in leading projects. Fragmentation of data-sharing approaches would limit the creation of seamless mobility, and lead to duplication of data storage, huge data inefficiencies, as well as a lack of coherence.

With increasing levels of complexity, we need different subsets of AI, such as machine learning and deep learning. When developing these technologies, it’s important to keep in mind that context-aware AI is of the essence. Up until recently, vehicle-related AI had a primary focus on the vehicle state. Now, as we’re gradually moving to higher levels of automation, it will be needed to incorporate information and predictions not only on the vehicle state, but also human (driver) state and system state (environment around the vehicle).

Another topic of increasing relevance is trustworthy and reliable AI. This will play a key role in supporting the public acceptance of AI-based connected, cooperative and automated mobility (CCAM) technologies and their market uptake, as well as in boosting the essential societal benefits (e.g. safety, emissions, inclusivity and different approaches to land use especially in dense urban areas).

Progressing beyond the current state of the art should thus firmly address the recommendations of the AI High Level Expert Group and its Guidelines for Trustworthy AI [1]. In brief, we should put effort into demystifying AI for CAD or CCAM to enable users to accept and embrace these new technologies - technologies which, I’d like to stress, need to reinforce the already good technologies of CAD. After all, AI is just a means to an end…

• 

  https://ec.europa.eu/digital-single-market/en/news/ethics-guidelines-trustworthy-ai

This article was originally published on the CAD Knowledge Base; the ARCADE project was funded by the European Union Horizon 2020 Work Programme

About the Author:
Margriet van Schijndel is program director responsible mobility, Eindhoven University of Technology (TU/e)