Autonomous vehicle technology crosses between both the physical and digital world. This runs into a problem in that humans currently test the physical and digital world separately
Seven years ago, Google launched its self-driving cars project. The very notion of this reminds us that autonomous vehicle technology has actually been ready for several years already. However, with traditional car manufacturers still running expensive Silicon Valley Research Centres, expanding self-driving fleets, forging partnerships, and investing billions into acquisitions, it’s still likely to be a long time before we see this technology going global. In fact, the perhaps-surprising reason for this delay in deployment is the limited testing and confidence in the technology. Autonomous vehicles highlight the importance of testing, since developers haven’t been able to do sufficient verification to give regulators and governments the confidence that the technology is ready.
Because of the lack of adequate testing and the resulting non-existence of confidence in the technology, anyone who does not work for a manufacturer is unlikely to see busy roads devoid of drivers for at least a decade. What is worrying is that this delay is costing over a million lives that could be prevented by avoidable road accidents. The delay in autonomous vehicles is quickly developing into a serious societal issue, rather than what would normally be a frustrating delay in new or ‘cool’ technology.
Autonomous vehicles highlight a few ways the current approach to testing is limited and won’t work in the future. For example, autonomous vehicles require millions of test scenarios, which cannot be completed through manually defining test-cases. For these test scenarios to be completed quickly and efficiently, they need to be intelligently auto-generated, otherwise the delay could be exponential.
In addition to the need for intelligent testing, the success and failure of autonomous vehicle testing cannot be rigidly defined. For example, there are numerous ways that an autonomous car can go round a corner that is ‘correct’, so the developers need to define various correct and incorrect ways to outline ‘success’ and ‘failure’ for the vehicles. Because of this, autonomous vehicle driving ‘failure’ has to be more heuristic based, rather than a basis that ‘exactly this must happen’.
Autonomous vehicle technology crosses between both the physical and digital world. This runs into a problem in that humans currently test the physical and digital world separately. For example, software testing is digitally tested, but the physical world is still not tested through automation. This consequently both limits and slows the testing process and doesn’t align testing efforts.
However, consider that when an object suddenly appears in front of a self-driving car travelling at speed, the safety system needs to immediately stop anything that may be distracting the person in the car (e.g. music or a phone call through the car system). Because of this, autonomous vehicle developers need to be able to test across both the physical and the digital world intelligently. In short, testing must be able to recognise the physical world and really be able to look at things from the user perspective, including looking at objects, looking at the screen, touching controls, and listening for sounds, to name a few examples.
The testing process in the autonomous vehicle industry is currently at a crossroads. It can continue the path of inadequate testing it is currently on, or it can bring together the physical and digital processes and test them using automation. This is the only option that will streamline the vehicle development. With autonomous vehicle technology already ready, speeding up the testing process will not only improve time-to-market, but will also, more importantly, drastically decrease both driver and pedestrian mortality.
Antony Edwards is CTO at TestPlant, leader in intelligent test automation. Antony studied computer engineering at the University of South Wales, Australia. He worked as a developer in Sydney before joining IBM Research in New York. Moving to London, Antony joined mobile operating system builder Symbian, moving from system architecture to eventually become a VP and member of the executive team. More recently he held the position of CTO with a major US on-line entertainment company.
If you enjoyed this, you may wish to view the following:
IBM patents cognitive systems for self-driving vehicles
The technology works in the event of a potential emergency and aims to be a safety measure that can contribute to accident prevention
Make way for autonomous vehicles: infrastructure considerations for urban environments by Michelle Caulfield-Harris
If the car manufactures and tech companies are to be believed, autonomous vehicles will be with us very soon, but key questions remain regarding the necessary infrastructure