As companies race to be among the first to put self-driving vehicles onto UK streets, several have been working on an essential first step: gathering data about those streets to allow the autonomous vehicles of the future to navigate with confidence.
One of these companies, the Academy of Robotics, is working on an autonomous delivery vehicle called Kar-Go that aims to reduce the last-mile costs of deliveries.
In our latest expert interview, we speak to William Sachiti, Founder and CEO of the Academy of Robotics, to discuss the company’s progress and discover the obstacles they face when digitally mapping our streets.
You've been involved in the industry for a while now. What first led you to work on autonomous vehicles?
I have always wanted to work in the space of automating last mile delivery. Originally, I wanted to do this with drones. To up my geek credentials, I decided to go to university and study Artificial Intelligence and robotics at Aberystwyth University. While there, it made sense that cars were a better route to market and thus the concept of Kar-Go was born.
“In the long run, driverless cars will save hundreds of thousands if not millions of lives.”
Kar-Go has recently started gathering data for autonomous vehicles on UK streets – where have you begun this process and why?
We have recently sent data-gathering cars into Surrey, Cardiff, Manchester, London, Brussels, Bologna, Munich, Berlin and Basel – essentially in cities of interest throughout Europe.
Their job is to go round a town to capture visual data in the form of video footage from up to 12 cameras with a combined 360-degree view around the car, as well as capturing feedback from sensors and infrared detectors.
This is all to gain a comprehensive understanding of the road environment and the roads’ users, particularly in residential areas. We will later use this data to teach the cars how to drive on these pre-driven roads.
What's the biggest challenge involved in the data-gathering process?
After we have gathered the data, we need people to sit and watch the footage, marking all points of interest such as highlighting where each car and object on a scene is frame by frame. We call this process annotation. It is very tedious.
How can the vehicle identify and differentiate between objects in the road – for example, pedestrians and street signs?
Much like a child is taught what objects are at school, we take images of similar scenes to roads where the car will drive. From these scenes we manually mark out what objects are i.e. each car, each person, etc. We call this annotation. Using a branch of computer science called machine learning, we apply the annotated data to an algorithm which now begins to compare images and learn the difference between a car, a pedestrian, a cyclist, the road, the sky, etc. After some time of doing this and us showing the computer more complex or harder-to-understand scenes, the algorithm in the computer eventually figures out the rest by applying what it has been taught and has learnt.
What mistakes might a self-driving vehicle make in interpreting what it sees, and how are you planning for this?
“We make the cars call home or request help from an operator if something is flagged as unusual.”
The risky mistakes are when an autonomous car sees a scene it does not understand or the sensors get a false positive caused by reflections or sun glaring into the lens. An example of this is a vehicle may not identify or know how to act when it sees an unclassified object such as a person on a Segway. Segways are not road legal but people may go on the road with one anyway. We make the cars call home or request help from an operator if something is flagged as unusual.
Once the autonomous vehicle has 'seen' and understood its environment, how does it predict what might happen next?
In the real world, if your smartphone were to slip from your fingers and start to fall, you know it will hit the ground. It is not going to stop all of a sudden and float or spontaneously shoot up. It falling and hitting the ground, to you, is a simple predictable action with an inevitable result. Similarly, the vehicle is able to see and identify pedestrians, cars, and bicycles etc. and then predict multiple realistic potential scenarios, taking action based on which potential scenarios are more likely to happen.
There's been a lot of media focus on the 'Trolley Problem' and the ethical choices that a driverless car might need to make. How significant an issue do you expect this to be?
We consider this a great philosophical topic of debate but one that doesn’t really make sense. How often have you or any person been faced with the same dilemma as the trolley problem in their day-to-day driving life? What would you do, and what would make your action or answer the correct one? It is easy to try to hold technology to some high moral standard and expect it to have answers to questions that we will struggle with ourselves. The fact to keep in mind is that, in the long run, driverless cars will save hundreds of thousands if not millions of lives.
“How often have you or any person been faced with the same dilemma as the trolley problem in their day-to-day driving life?”
A number of UK companies have now begun gathering data for driverless car projects. Is there a point when it will become necessary or desirable for companies to work together on data-gathering, or to develop common standards for this data?
I think competition is good and it will force us all to continue to innovate. There is already a lot of shared data in the autonomous car space. These open data sets are available for anyone to download. By gathering our own data, it allows us to hyper-focus on the exact problem we are trying to solve.
Finally, what are you personally looking forward to most about the arrival of self-driving cars?
I look forward to seeing cost reductions for everything that involves transport. For me, I would like to see a world where moving people and products anywhere costs next to nothing.
About the expert
William Sachiti, Founder and CEO, the Academy of Robotics
William Sachiti is the Founder and CEO of the Academy of Robotics, a UK company specialising in creating ground-breaking robotics technologies such as autonomous vehicles. Their first commercial solution, Kar-Go, is an autonomous delivery vehicle that aims to vastly reduce the last-mile costs associated with deliveries.
Find out more about the Academy of Robotics
We speak to Rita Excell, Executive Director of the Australia and New Zealand Driverless Vehicle Initiative, to discuss the impact of driverless vheicles on industry jobs and the risks of a ‘driverless divide’.