Autonomous Vehicle Infrastructure – Are we closer than we think?

The problem that came to mind

While much of the discourse around autonomous vehicles (AVs) has centred on when the vehicles themselves will be technologically ready for deployment, far less attention has been paid to the readiness of the road infrastructure they will operate on. At Nota Bene we decided to carry out our own research, seeking to address that gap by examining how existing roads, using Sheffield as a case study, might already meet, or fall short of, the requirements for AV integration.

Our Thoughts

By applying a scoring framework based on physical and digital criteria, the study aims to provide insight into where AVs could be safely and effectively introduced, contributing to a more grounded understanding of practical AV deployment in urban environments.

In recent years, the introduction of autonomous vehicles (AVs) on public highways has become an increasingly imminent prospect. While the vehicle technology is being tested and may be almost ready, there has been limited focus on the requirements that the existing infrastructure may need to meet to be ready to host AVs. However, many sections of the network may be closer to being AV-ready than we think. 

This article draws on existing research to assess suitability and create a scoring system for the existing road network to assess the suitability of the major road network alongside all other supplementary A-roads for hosting AVs, using Sheffield as an example. 

Automation in vehicles is classified into 5 ‘levels’ with level 1 representing a single automated feature such as cruise control, all the way through to levels 4 & 5 where a vehicle can operate without the need for human intervention.  

One of the most important reasons for the need to classify the suitability of the existing infrastructure is to ensure that there can be safe interactions between human driven vehicles (HDVs) and AVs as well as with pedestrians and cyclists, as these interactions are much more unpredictable than interactions between multiple AVs. 

Opinions on what the requirements of the road infrastructure to host AVs will be are mixed, with some arguing that the existing infrastructure will be suitable enough as it is, although it is worth noting that many who hold these opinions work in the automotive industry. On the other hand, many industry experts believe that initial rollout would be within a defined geofenced area. 

Many of the requirements for AV infrastructure are simple enough to implement but may require a higher level of road upkeep such as not having potholes, more regularly maintaining visible, reflective road markings and ensuring that traffic regulations are designed in a machine-readable format.

The potential requirement for higher standards of maintenance to ensure AV operate safely will have an impact on local authorities’ road maintenance budgets.  

Other potential requirements relate less to physical aspects of the infrastructure and relate to digitised aspects such as mobile signal, the availability of high-definition mapping or the presence of variable message systems, which can be used to digitally transmit live data and information to AVs. 

Building on the existing research, a weighted scoring system was created using the likely requirements of the road infrastructure to host AVs, a list of the existing main roads in Sheffield, alongside other A-roads were collated and manually scored using Google Street View to give each road a suitability score for hosting AVs. This data was then mapped using GIS to illustrate where the most suitable roads were located and where any patterns might appear such as clusters of roads with a high suitability score located close to each other. 

The suitability map of the main roads shows a high density of roads which may already be suitable or require a small number of upgrades to be suitable, this could perhaps feed into a network of city centre ‘robotaxis’ which serve the inner ring road and a limited number of through roads. Roads on the periphery of the city tended to score lower due to fewer lanes, less provisions for vulnerable road users and poorer maintenance. 

One factor which was not considered in this study was the storage and recharging of the vehicles. If the initial rollout of AVs is for use as taxis, there will need to be sufficient space for their storage when not in use, likely needing to be along one of the suitable roads. One potential solution for this would be to make use of underutilised parking spaces. Additionally, the same issue would need to be addressed for recharging the vehicles.

Having gone through this process, it is likely that when rolling out to other towns and cities that an AI model could be used to automate the scoring of the road networks. However, it is important that a human touch is maintained as part of the process for cross-referencing any mistakes which the model may make.  

It may be interesting for future studies of the same nature to expand the scope to a wider area, such as the Major Road Network across an entire region to assess if a well-connected autonomous vehicle network could be feasibly established.  

Ensuring that autonomous vehicles can safely navigate a city or region using the infrastructure will be key to ensuring an effective roll out without hindering public perceptions of their safety.