The companies have announced they will be collecting data and accelerating autonomous vehicle development with Kar-go.
The deal gives AI startup, Academy of Robotics, access to one of the broadest sets of varied data on UK roads to train its Kar-go self-driving operating systems and provides Eurovia UK with ‘smart’ solutions for improved highways management.
Academy of Robotics and Eurovia UK today announced plans for a national partnership to trial the use of Kar-go’s autonomous vehicle technology to support urban and rural highway management and to accelerate the “training” of Kar-go to navigate on unmarked roads in the UK.
Academy of Robotics is the startup behind Kar-go, Europe’s first road-based autonomous delivery vehicle. Originally developed to dramatically reduce both the financial and environmental costs of last-mile parcel delivery, Eurovia UK recognised its potential for use in infrastructure and urban development work and is keen to test the Kar-go technology to automate the delivery of small plant equipment, tools, materials and other components to and from a highway work site as well as the potential use of data collected by Kar-go as it travels, to determine the condition of roads.
Eurovia UK, a leading specialist in operating, maintaining and improving road, highway and public realm infrastructure will help Academy of Robotics to scale their training of autonomous vehicles. This will be enabled by accessing digital camera data gathered by Eurovia’s fleet in the UK, which covers over 50,000 kms of UK roads.
The trial will initially focus on the UK, but as a global business operating in 15 countries, the deal offers significant potential for international expansion in Europe and North America.
Understanding hazards – shadows vs. potholes
Through their unique approach to colour perception, Academy of Robotics have been able to address the challenge of environmental variability and ‘noise’ caused by shadows and bright reflections meaning that Kar-go’s operating system is already able to understand the difference between features such as cracks, puddles, potholes and shadows.
William Sachiti, CEO and Co-Founder of Academy of Robotics commented, “To date, most autonomous vehicle training and testing has taken place on well-marked roads or specially designated test centres, but these areas rarely reflect real-world conditions. We believe that training our vehicles to operate on the widest range of real-world conditions is critical to preparing them for the unpredictable elements vehicles may face in the ‘real world’. The AI can process the data at immense speeds, so one of the biggest challenges holding back this technology is the ability to get hold of enough data on diverse conditions to train vehicles at scale.”
The technology Academy of Robotics has developed is able to detect not only the potential hazards in the path such as the edge of a road in snowy conditions, but also the likely causes of deterioration on road surfaces. Reflecting Eurovia UK’s commitment to improving road surface conditions, this collaboration will ultimately offer a proactive, ‘smart’ approach to detection of both highway defects and hazard prevention. As Kar-go vehicles are electric, the partnership also supports Eurovia UK’s work in investing in projects that support green growth: developing a greener fleet and reducing CO2 emissions.
Scott Wardrop, Chief Executive of Eurovia UK said, “We were impressed by the sophistication of the technology developed by Academy of Robotics and the professionalism of such a small startup team. We have reviewed a number of autonomous vehicle solutions, but a critical component for us in developing this partnership was the technology’s ability to manage the complexity of recognising different road surfaces and their absolute commitment to sustainable innovation – investing in the future. As a Sussex-based company, we are also proud to support a local start-up that has built the first Kar-go in Henfield.”
Designed specifically for making deliveries, the Kar-go vehicle technology uses a “terrain-training” approach. Much like a delivery driver, the vehicles need to become familiar with the catchment area in which they are driving and they are then able to recognise and respond to hazards and variables they encounter on their route.
The first stage involves mapping an area and then training the technology using AI based on advanced neural network on how to respond to the elements that have been mapped. The more data the system can access, the faster and more accurate this technology becomes. The mapping element can be done by simply applying cameras to a car and then recording footage of driving on a road. The car’s camera footage can then be used to map an area and train the autonomous vehicle to drive in the area which the camera has seen.