You may have driven down the motorway on a clear summer's day and seen a sign reading "FOG"!
To prevent this kind of misinformation, emerging transport applications, like drones and driverless cars, will require navigation performance that cannot be delivered through one source alone. In the aviation world we use GPS and SBAS to provide navigation guidance across flight phases. This is safe, because we know the level of performances and, crucially, we can detect if it has failed. We also have a lot of space to play with compared to applications on the ground or even drones.
A self-driving car needs to be certain of where it is within much tighter limits than, for example, the 0.1NM required by RNP0.1. How will this happen? High accuracy maps are essential, and fusing information from many sources is needed, to name a few: GPS (and other GNSS like Europe's Galileo), inertial sensors, radar and LIDAR, and cameras enabling computer vision. Signs that read "FOG" on a clear summer's day will no longer be acceptable, because we must be able to trust them!
The challenge that arises is not one of technology. Although achieving the performance required is no mean feat, we are already well on the way to delivering it. The challenge lies in testing and proving the systems. In aviation we can model the entire system and satisfy ourselves that it is as safe as practicable and meets our requirements. However, systems that incorporate more data and ever more sophisticated processing, cannot be modelled sufficiently to satisfy ourselves of their safety. Google can work out how a specific individual search result was generated but ask them how any/all search results will be generated and you won't get an answer.
We can test the components of the system just fine, but not all the possibilities when you plug them together. Laboratory testing, whilst allowing us to adopt a systematic approach (isolating variables and testing failure modes) won't be able to cover the full gamut of possibilities (for the same reason we cannot model the system). Real world testing will allow us to explore the subtleties of the system in use (i.e. how does the user interpret a given indication?), but cannot be systematic, unless you can control the weather, external systems (like those road signs) and other people.
The answers will involve systematic laboratory testing of components, and test campaigns of integrated products in the real world. The UK has some great facilities for these, and new procedures and regulations for approving systems. But given the risks and variables involved, it will be a very brave person who signs off the first entirely self-driving car with Level 5 automation, where the driver becomes the passenger, and the steering wheel is an optional extra!
Three costly mistakes in ATM systems upgrade projects
Sustainable aviation: what are today’s most promising solutions?
Why you need ‘quick wins’ – even in aviation
Aviation emissions – virtue signalling or real action?
Proactive vs reactive defence in aviation cyber security
Behind the scenes at an airport near you?
Data centres – deal or no deal?
The role of Human Factors in de-risking COTS implementations