Oxford University and Oxfordshire are at the centre of the autonomous vehicle revolution. Professor Paul Newman, co-founder of the University spinout company Oxbotica, talks disruptive software, job creation, and why there’s more to automation than driverless cars.
Stuart Gillespie: Oxbotica is approaching its five-year anniversary: how do you reflect on the journey so far?
Paul Newman: It is extraordinary to reflect on how much has been accomplished since we set out on this journey. Oxbotica was founded on a recognition that software is the main force of disruption across industries. It has been an enormous joy to see that come to life with the application of our technology in a wide range of environments worldwide, from city centres, ports and airports to mines and forests.
Everything has moved at a tremendous pace. In the last six months alone, we secured an investment round, formed a partnership with Addison Lee, and have had autonomous trials running on public roads in London and Oxford. In the coming months, we’ll also be going public with a number of other strategic partnerships.
From its inception, Oxbotica has worked closely with both local and central government, as well as transport authorities in Oxford and London. We are creating amazing jobs here in Oxfordshire and, via that, exporting software to some pretty extraordinary places. By the end of 2018, our software was running in four continents.
All of this comes down to the sheer talent and determination of the world-class team we have built across every area of the business, and it’s a team that is always growing because we are scaling up so rapidly on a global level.
SG: How do your software systems work? What applications do they enable?
PN: Selenium is our autonomous control system, effectively functioning as the brain of an autonomous vehicle. It is a suite of software that combines data from vehicle sensors to drive autonomous vehicles. It answers the questions ‘where am I, what’s around me, what do I do?’ Selenium identifies obstacles – such as people and other vehicles – before calculating a safe and efficient route.
Caesium is our cloud-based fleet management system that schedules and co-ordinates autonomous vehicles without human intervention. Enabling the exchange of data between vehicles, it optimises routes and ensures that fleet managers know exactly where all vehicles are and when they will arrive. It’s also the mouth of the machine learning data acquisition system, and we have some interesting plays there which cunningly leverage Selenium’s unique capabilities.
What I enjoy most technically is dealing with the diversity and complexity of the technical approaches needed to address the autonomous vehicle problem. The term AI doesn’t really do it justice. Someone put it to me that we have an embarrassment of technical riches, and that’s true because we have such a richness in underlying techniques from every corner of the AI spectrum.
SG: Driverless cars: where are we with this technology, and what does the future look like?
PN: The rate of progress has been so high in the last year alone that it’s left some people confused about what the near-term future holds. It won’t be an overnight transformation of our cities in which driverless cars will pick people up and take them to any location at any time.
Within the next two years, we’ll begin to see the deployment of autonomous cars in geofenced areas – predefined areas in which they will operate as a service. This can serve a wide variety of purposes, such as transporting people in a taxi service, acting as a courier for legal documents, or a delivery service for groceries in areas around our cities.
In the longer term, of course, autonomous cars will become ubiquitous and form a core part of transport infrastructures all around the world. It is hugely encouraging to see the UK government develop plans to enable advanced trials of autonomous vehicles, because we always want to see policy keep pace with the amazing innovation taking place.
SG: Contrary to the image often presented, there are plenty of other uses of autonomous technology beyond driverless cars…
PN: There is huge appetite across a range of industries, from mining and agriculture to retail and ports, for autonomous vehicle software to radically enhance safety and efficiency levels. But those purpose-built vehicles require heavy investment and are not easily replaceable, which is why it is only a software solution that can deliver the promise of autonomy to them.
Software scales in a way that physical objects don’t. When it comes to autonomous vehicles, effort and bottom-line cost are essentially independent of the number of vehicles in which the software is deployed. The returns are also proportional to the fleet size, so you’re buying one piece of software to be used on 10,000 vehicles rather than 10,000 self-driving vehicles.
We built Oxbotica’s autonomous software to enable any vehicle to operate in any environment, without relying on external systems such as GPS or third-party maps. The stack was built from the ground up, component by component, and targeted at any vehicle. We didn’t choose a particular sensor modality – vision, lidar and radar all have their roles and can be used solely or jointly depending on application type. It’s a complicated business technically, but we are nailing it.
All of this has made software-led autonomy the most credible and realistic path forward for all industries.
SG: What’s next for Oxbotica?
PN: Oxbotica is in a period of rapid growth – we’re growing the team, and we’re growing our presence in a fascinating set of industries in global markets. We are intent on building an extraordinary AI company in the UK – one that changes the way vehicles operate. As part of that we will continue to grow, form deep partnerships, and write some pretty amazing code.
It’s incredible that something with zero mass can change so much. Bits to move atoms – that’s us.