An extra gear

The transport sector is responsible for almost a quarter of energy-related gas emissions. The use of Artificial Intelligence can produce the radical innovation needed to speed up the transition to a more sustainable system.

by Aidan O’Sullivan
14 min read
by Aidan O’Sullivan
14 min read

The transport sector and carbon emissions

Transportation has been fundamental to human society and economies throughout the ages, enabling trade through the movement of goods and materials; and the transfer of skills, ideas and the growth of communities through the spread of people. The democratisation of transport and increased access to travel has been key to enabling ever more people to venture ever further in the modern globalised era in which the world has become ‘smaller’. However, the current levels of mobility are an unsustainable threat to our future, the global transport sector now accounts for 23% of energy related carbon emissions and in some countries such as the UK is the largest contributing sector to carbon emissions. Despite efforts, progress towards more sustainable mobility has been slow with the world bank stating:

"The world is off track to achieving sustainable mobility. The growing demand for moving people and goods is increasingly met at the expense of future generations"  José Luis Irigoyen, Senior Director of the Transport & ICT Global Practice at the World Bank.

Indeed, the contribution of the transport sector to global emissions is likely to get worse with global air traffic passenger demand currently growing at an annual rate of 7%, averaged over the last five years, equivalent to a doubling rate of total passenger demand of 10 years. Furthermore, there is a need to improve access to transport in developing countries, about 450 million people in Africa, more than 70% of its total rural population, are estimated to have been left unconnected to transport, and emissions could grow by 40% by 2040. Global Mobility Report 2017. Additionally, in the UK, the Committee for Climate Change has criticised the sector for failing to make anything like the progress demonstrated by the power industry in responding to the threat of climate change.

Radical change is needed

It’s clear from this picture that radical change is needed in transport and it is critical that a sustainable mobility revolution is launched to make the kind of progress required in a sector where emissions reduction is challenging. There are reasons to be optimistic though and believe that this sector can be transformed. In the last decade, digital technology has enabled the emergence of innovative new business models, such as the ride sharing platform Uber, that have brought major disruption to the transport sector, challenging established norms and creating multi-billion dollar companies at breakneck speed. This is the type of radical change and innovation needed. Fully decarbonising the transport industry presents major challenges, particularly in sectors like aviation, however just as progress in digital technology brought disruption, recent progress in the fields of big data and Artificial Intelligence has the potential to solve many of the sustainability challenges and enable the sustainable mobility revolution. Indeed, one of the key challenges that AI can help with is the sheer complexity of transport systems.

The global trend of urbanisation with ever more people drawn to live in cities, places the urban environment on the frontline of the climate battle as a dense concentrated source of emissions where air quality has major health implications. However, this also presents the opportunity and motivation to make significant progress by developing more sustainable urban transport infrastructure. Urban transport systems are made up of numerous competing and interacting modes of travel with different characteristics in terms of levels of service, cost, frequency and speed. Coupled with this is the heterogeneity of the residents of the city and their transport needs, the spatial distribution of the city in terms of land use mix and accessibility. The complexity of such a system presents a major challenge to methodology used to inform decision making and trying to assess the best allocation of resources and their impacts, with the non-linear interactions and feedbacks highly challenging to model mathematically. One solution to this is found in the field of agent-based modelling where the complexity of a system is modelled by faithfully reproducing the components of the system and allowing them to interact intelligently. Artificially Intelligent agents that capture with high fidelity the everyday decision making of individuals as they travel through the city would be a huge boon to understanding how urban transport systems work and informing how best to transform them to become more sustainable. While these modelling tools have been developed previously the sheer scale of modelling a city in high fidelity is a gargantuan task and previous attempts have required approximations and heuristics. However recent progress in neural networks and scaling these methods, holds the promise of much more sophisticated agents. In addition to this mobile phone data is increasingly being used as an invaluable resource to understand how people move through a city and the combination of these two technologies could produce a type of ‘virtual city’ environment of sufficient realism that would enable the development of bespoke solutions for specific city problems. For example based on demographics and culture the type of transport system required in a young Asian city may be completely different to that required in a slow growing European city.

The potential of AI

While AI can help model cities virtually to make strategic investments in infrastructure it is also enabling new forms of transport in flexible on demand services. More flexible use of resources across a city involves large scale decision making that AI enables in terms of managing a complex portfolio of distributed assets. It also requires ‘context aware’ decision making and an understanding of the factors driving travel demand. An example of this is rare events, rather than over engineer the transport system to cope with rare peak events such as a huge concert or international sporting event, the ability to shift capacity and adapt frequency in expectation of these events would improve the operation of urban public transport systems incentivising greater ridership and reduction in private transport emissions. Natural Language Processing, the AI field associated with developing algorithms that can understand human language has made great progress in translation and recognition however we are still far from systems that `understand’ the meaning of language. Further progress in this area could enable an advanced transport oracle that takes in information about the world from the news, ticket websites, weather forecasts, music charts, sports tables and based on this information about an event is able to reconfigure the transport system to meet travel demand needs by making better use of existing resources. Furthermore, in the developing world flexible on demand services driven by data have the potential to help ‘skip a generation’, just as we’ve seen with mobile technology. For example in a Data 4 Development project sponsored by Orange Telecom researchers at IBM redrew the bus network based using mobile phone data and found they were able to reduce travel times by 10%. Technology like this combined with AI can allow for bus networks that reconfigure themselves in real time based on the location information of users.

It is impossible to discuss AI and transport without considering autonomy. More than any other technology autonomous vehicles have the potential to bring about radical transformation in this sector. Currently a number of methods are being trialled such as LIDAR based navigation however as AI and Computer Vision progress it seems inevitable that the future system will be designed around cheap camera sensors that use sophisticated AI algorithms to ‘see’ the world enabling the roll out of self-driving vehicles that can exceed human performance in terms of safety and reliability. Clearly these vehicles need to be electrically powered in order to have sustainability benefits, however autonomy enables a new form of business model Autonomous Mobility On Demand. Similar to ride-sharing under this paradigm car ownership would disappear and either a government or company supply a fleet of autonomous vehicles which can be hailed by passengers through their mobile phones to take them to their desired locations. Simulation studies have shown that it’s possible to serve significant numbers of passengers with much fewer vehicles on the road through this type of shared resourcing, which reduces congestion and consumption of resources for a more sustainable system. This business model could also be a key enabling factor in the financing of Electric Vehicles which have high upfront costs but much lower operating costs, so operating this asset in an AMOD scheme would be an excellent means of maximising revenue.

Autonomy in the aviation sector

Autonomy is not just limited to road vehicles and it may in fact be even more transformative in the aviation sector. This sector of transport faces considerable challenges in achieving zero carbon aviation, with the energy density of petroleum products very challenging to match. While electric power boasts greater efficiency in conversion of stored energy battery technology will still need to evolve considerably for it to be viable to consider powering current aircraft such as an A320 or Boeing 747. Autonomy may allow us to rethink aviation. Airbus is developing a new type of vehicle through its Vahana project, an electrically powered vertical take-off and landing autonomous air taxi. Since the vehicle is autonomous it does not face the costs associated with a pilot that modern aircraft incur and which clearly scale with the number of aircraft. This could enable a different business model for air travel with the mass transport of 180+ passengers in a single rigorously scheduled vehicle becoming the way of the past and the future being much greater numbers of far smaller electrically powered craft transporting passengers autonomously. This is similar to developments we are seeing in freight where companies like Amazon are funding research into replacing large delivery trucks with autonomous flying delivery drones. Clearly navigating a city safely requires a level of AI that currently doesn’t exist but progress in this area as well as in swarm intelligence could allow for self-organising fleets of delivery drones that work together to deliver packages more efficiently and sustainably replacing fleets of delivery trucks. In the shorter term freight companies are already using AI and data to make better use of resources and the US parcel delivery company UPS has demonstrated impressive results from investing heavily in an analytics programme known as ORION (On-road Integrated Optimization and Navigation) which is the firm’s fleet management system. Optimising the routing of trucks and delivery of packages is estimated to save UPS 100 million miles per year or a reduction of 10 million gallons of fuel ref. While this has obvious revenue benefits it also aids sustainability reducing carbon emissions by 100,000 tonnes.

Towards “smart” batteries

Battery technology is set to be a cornerstone of the move to a more sustainable transport system with electric vehicles charged by renewably generated electricity representing a zero-carbon solution to our transport needs. While the chemistry places a fundamental limit on what is achievable in terms of some aspects of battery performance AI can still play a role in intelligent battery management systems which optimise temperature and performance within the cells of the battery to extend the lifetime thereby reducing capital costs associated with the technology. Capital costs are one of the chief drawbacks of electric vehicles which are far cheaper to run but require greater upfront investment. On a more radical level AI algorithms can be used to test millions of different combinations of battery chemistries in simulation to develop the next generation of battery technology. This is akin to computational drug discovery, where the space of possible combinations is vast and intelligent algorithms such as genetic evolution methods are required to search the space efficiently as even with today's sophisticated computing infrastructure a brute force search is simply infeasible.

To conclude it is vital that a sustainable mobility revolution is launched and soon. Progress in AI allowing for ‘narrow AI’, algorithms that can make complex decisions to solve specific tasks in uncertain and dynamic environments, have been able to demonstrate superhuman performance and are a vital technology in enabling and supporting this revolution. Application of this technology in the transport sector can produce the radical innovation desperately needed to accelerate the transition to a more sustainable transport system. Key applications that are actively being researched and have the potential to solve some of the particular challenges associated with the transport sector are; AI virtual city models for transport planning, algorithms that enable flexible on demand public transport services that are context aware and adaptive to the state of the world, autonomous vehicle technology in both road and air vehicles, advanced drone delivery technology replacing current freight fleets and AI algorithms that can enhance current battery technology and aid the development of new technology. However these examples are just the tip of the iceberg and many more radical applications of AI will be needed in the sector in the coming years.ore.


The author: Aidan O’Sullivan

He is a lecturer at the UCL Energy Institute, University College London. He is course director for the MSc. in Energy Systems and Data Analytics and Head of the AI and Energy research group.