The number of high rise buildings has risen by 650% in the last five years, creating “urban canyons” which cause GPS signal disruption. Autonomous vehicles could face signal black-out caused by tall skylines found in cities across the world.
Interruption of GPS signals, through connection drops or multi-path degradation, has potential safety and operational implications for autonomous vehicle systems that rely heavily on satellite-based navigation. This is where Oxbotica’s pioneering technology comes to aid. The technology is able to work independently of any external infrastructure, allowing continuous localisation and safe control of its vehicles – even without a GPS signal.
Let’s explore the nuances of the innovation with Ben Upcroft, VP of Technology, Oxbotica, UK.
What bothers you about the use of GPS in autonomous vehicles?
Across the Autonomous Vehicle (AV) industry, there is an over-reliance on GPS to establish the position of the vehicle relative to its surroundings. This is arguably one of the most safety-critical questions that a self-driving vehicle must answer – where am I? Interruption of GPS signals can cause serious safety issues with autonomous vehicle systems. With high rise buildings becoming a common phenomenon, this kind of interruption is likely to occur more often, and that’s why we need an alternative to GPS for AVs.
Tell us about the new technology invented by Oxbotica.
Oxbotica hasn’t just developed autonomy for the car, we’ve developed autonomy completely independent of any one vehicle or use case – we call it Universal Autonomy. It means that the same autonomy software can be used on any vehicle in any environment. If a GPS signal is available, of course we will utilise it, but we never rely on it.
The company’s system, called Selenium, ingests data from cameras, lidars, radars, and a number of other sensors if available. Using the sensor information, Selenium establishes where the vehicle “is”, its environment, and thus how it should move. Selenium provides an on-vehicle suite of software requiring very modest compute hardware, delivering full autonomy to any land-based vehicle.
Selenium is not dependent on any single sensor modality (it leverages independence and redundancy in sensing) and because it makes no assumptions about a vehicle’s working environment, it is able to operate in a vast range of settings, on any vehicle platform, under any conditions. It can use radar, laser or vision sensing and as these technologies evolve, offers unparalleled opportunities for future development.
Is this the beginning of a ‘no-GPS’ era for AVs?
Our software is already used in quarries and mines, and in ports and logistics hubs, and works just as well underground or indoors where there is no GPS. It does not even need HD-Maps (although this can still be utilised, if available). As a result, the system can be tailored to work in any environment, indoors or outdoors, on a large scale. I don’t think it will spark a complete departure from GPS – and neither should it – but we don’t believe that AVs should ever be wholly reliant on GPS.
How has been the acceptance of the technology so far?
Selenium, and indeed our approach to autonomy has been received very positively. We work with customers and partners at varying stages of development, with rapid deployments due to our domain agnostic approach. Through industry, we have deployed and operated our autonomy across multiple types of vehicles, across four continents, from underground mining, to city commuting, to warehouses – all with the same code. Our technology is continuously learning in each of these domains and sharing those learnings across the domains. Ever increasing our robustness and reliability.
How will it impact the geospatial industry?
We believe that our work complements the geospatial industry and accelerates it. Ultimately, what we do is not dissimilar, we are both in the business of gathering and analyzing data that allows us to build a picture of a particular location on the Earth’s surface. We believe that the richer the information you can attribute to that location – through radar, lidar, and computer vision technology – the better AVs will perform.