Inria unveils its White Paper on Autonomous and connected vehicles
Robotics & Intelligent Transportation Systems - © Inria / Photo H. Raguet
Following the publication of its first White Paper on Artificial Intelligence, Inria unveils a second document, this time on autonomous and connected vehicles, in which it identifies and highlights all the issues surrounding the major advances and innovations in this field. What are the technological and scientific challenges, the economic, social and environmental issues, and the legal and ethical implications? This White Paper is a reference text that evaluates, questions and looks ahead.
"Supplementing the national strategy that was presented today by Anne-Marie Idrac, Senior Official for the national strategy on the development of autonomous vehicles, this White Paper on autonomous and connected vehicles demonstrates the importance of scientific research to enable progress, particularly in terms of the software, cybersecurity, human-machine interface, modelling, simulation and artificial intelligence challenges. The topic is in fact extremely complex, firstly because it does not define a single problem but rather encompasses an array of different kinds of technological and scientific issues, and secondly because future advances will not come about by fine-tuning current techniques and processes. We will need to invent new concepts and new models for the design, manufacture and use of vehicles,” commented Peter Sturm, Inria’s Deputy Scientific Director in charge of research into perception, cognition and interaction.
Autonomous vehicles: a radical shift
The world of mobility is undergoing profound changes. Autonomous vehicles will not only mean completely new modes of transport but, more importantly, will have an impact on society in the future – in terms of security, environment, urban development, etc. – and also on the automotive industry itself, with new vehicles in which artificial intelligence will play a crucial role. To gain a better understanding of these radical changes, we first need to understand the specific features of an autonomous and connected vehicle, the feats of technology the industry players will have to achieve and the collaborations between public and private research that will need to be developed. For many years now, Inria has had teams of researchers from different scientific fields working on this topic. Many different aspects of research are concerned, including mathematical modelling, robotic techniques, human-machine interfaces, telecommunications technology, networks, cybersecurity, machine learning, decision-making tools, reasoning in uncertainty, and more… All these research themes are touched on in Inria’s White Paper.
Four strategic questions about autonomous and connected vehicles
No manufacturer believes it can escape the move towards autonomous vehicles. However, before they can be deployed on a large scale, four outstanding strategic questions will have to be answered. Inria asks, and expands upon, these questions in its White Paper:
- How do you ensure system security and reliable communications?
- Can an autonomous vehicle understand its environment?
- Can automatic steering be intelligent?
- How do you validate such a complex system?
Multiple technological and scientific challenges
The design and development of autonomous and connected vehicles pose considerable technological and scientific challenges, which include:
- Autonomous navigation
For a vehicle to navigate autonomously, without a driver, it must be able to detect its environment, analyse and interpret the data it receives and make decisions. Consequently, the challenges of multi-sensorial perception (sensors, scenario interpretation) and autonomous decision-making (planning itineraries, manoeuvres and journeys, and vehicle-user and user interactions) are enormous. High-level control-command methods are also under consideration to ensure the vehicle can offer optimal stability and passenger comfort (limits on longitudinal and lateral acceleration), avoid or correct skids and control wheel slip.
- Incorporating software and operational safety
Incorporating software into a highly automated vehicle is therefore particularly challenging. Considerable research is still required in several areas, including embedded architecture design, formal programme validation, hardware and software optimisation, resilience, error tolerance, uncertainty management and the security of the physical and computerised systems.
- Telecommunications and cybersecurity
The deployment of autonomous and connected vehicles will require the telecommunications networks to be considerably adapted and made secure. The massive increase in the number of communicating agents will have a huge impact on the occupancy of already highly saturated communications channels. Road safety requires fast and reliable exchange of information between neighbouring vehicles in all traffic conditions. The challenge is therefore to create reliable wireless methods of communication that work in very dense traffic.
With vulnerability to external attacks and the integrity and confidentiality of the information circulating in the networks more important than ever, cybersecurity is becoming a priority for the transport industry. There are already protection systems on the market, but the challenge now is to incorporate them into the vehicle development process.
- Big Data
The systems will produce vast quantities of different types of data: an autonomous and connected vehicle could generate up to 1 GB of data per second. Considering that hundreds of thousands of connected vehicles are expected to come into circulation, all this data will have to be stored, managed and processed automatically by processing centres, operators or users. It has not yet been decided how this big data will be stored and processed, which will present considerable financial and technological challenges.
Digital simulation validation is a particularly difficult challenge. Even if every sub-system could be validated individually, how do you validate the complex embedded system when in actual use on the road? You would need to simulate the workings of all the technological building blocks simultaneously, incorporating the complexity of on-the-road scenarios and the behaviour of other drivers, bearing in mind that there will be vast numbers of autonomous vehicles driving around at the same time. There is currently no simulator that is suitable for all the functions of an autonomous and connected vehicle.
- Modelling the key systems: road traffic and fleet management
With a large number of connected vehicles on the road, the traffic model will need to be fed with the data collected by the vehicles, which will act as moving sensors. Modelling also has an important part to play in managing fleets of, for example, shared vehicles. Here, one of the main issues is re-allocating the vehicles in an increasingly busy and dense urban network.
- Human-machine interfaces
With new functions incorporated into vehicles, and changes in driver behaviour, specific human-machine interfaces will have to be designed for autonomous and connected vehicles. These will be for occupants of the vehicle and also other road users with whom interaction will now be required. The teams that will work on these interfaces will have to investigate one very important point: juggling automatic steering and driver steering, depending on context and situation.
A societal, legal and ethical impact
An autonomous and connected vehicle will not only be a system of great technological complexity; its large-scale deployment will also bring about profound changes in how society operates, which will affect how cities are organised and transport is used, and the civil liability of mobility manufacturers and operators. The arrival of autonomous vehicles thus raises many legal and ethical questions.
In its White Paper, Inria also responds to the much-asked question about timescale.
"The first automated transport systems, on private or controlled-access sites, should start appearing in 2025. This is also when we expect to see autonomous vehicles on the motorways, provided the infrastructure has been adapted (dedicated lanes, for example). It will be 2040 before we see fully autonomous vehicles, requiring no human intervention, in peri-urban areas and being tested in cities," said Fawzi Nashashibi, head of Inria’s RITS project team. "However, maturity of the technology is not the only hurdle to deployment of these vehicles. It will depend largely on political decisions (investment, regulations, etc.) and land-use planning strategies."
These articles could interest you:
- Read the White Paper on Autonomous and connected vehicles ! (in French)