Automated cars and data

Everyone’s talking about automated cars and how they will make it cheaper and easier for us to get from place to place. As well as helping us travel they will change our cities by freeing up space, save lives by reducing the number of driving accidents and lead to the loss of millions of driving jobs with the associated impact on people and communities.

If you’re reading this I bet you’ve heard this talk. If you live in one of the test areas in America, UK, China and etcetera you may even have seen trials. There are skeptics, but I think that people will be able to gradually build ever safer and more automated cars. Once they do many people will choose to use them. Change is coming.

More Autonomous Cars Coming to Public Roads in 2016 Copyright © 2016 ENGINEERING.com

Making it easier and cheaper to move around, changing cities, saving lives, removing a type of job are complex things. There are many more secondary effects. Our policymakers need to consider the risks and benefits to help us get to a better society that includes automated cars and benefits everyone.

But I’m not seeing enough discussion of one important aspect of automated cars: data, and how security, privacy and openness can increase its impact.

Automated cars collect a lot of data

As well as transporting people and parcels automated cars will collect vast amounts of data. A human driver needs to look around to see street signs, the weather or cyclists. Similarly automated cars will need to collect data to make driving decisions.

http://dataconomy.com/how-data-science-is-driving-the-driverless-car/

Automated cars collect a lot of data. A PhD student recently calculated that a modern car already generates 25Gb of data an hour. In 2013 it was reported that Google’s automated car generates 750Mb of data a second. Earlier this year Comma.ai, a company that was working on automated cars released 80Gb of data generated during 7 1/4 hours of driving.

This data includes such things as the car location, maps and video footage of the surrounding area, information about nearby traffic, accidents, weather information, the route of the car and information about any passengers or parcels that that are in the car.

That’s a lot of data, how do we get most value from it?

Security and privacy

The security of this data clearly needs to be considered. We need to protect the data collected by the car and the data that the car needs to be able to get to do its job. Car hacking is a real risk whilst an automated car is likely to be more dependent on access to data than a car driven by a human. Data is already an under-recognised piece of critical national infrastructure, automated cars will only increase the need to strengthen it.

Silly Wired. Nexar, like any camera, isn’t just collecting your data it’s also collecting data about other people.

Privacy will also be an important consideration. If automated cars mishandle personal data about the people travelling in them or the people and things seen by their video cameras then some people will be damaged while other people may lose trust and choose not to use the cars.

Some of these issues will be explored by smartphone apps, like Nexar, that use the smartphone’s camera and microphone to collect data about car drivers, passengers, pedestrians and other cars.

But automated cars will collect far more data than a smartphone camera.

Automated cars will use data collected by other cars and people

Automated car manufacturers and policymakers should be thinking about security by design, privacy by design and how openness can help build the trust that will be needed to get the most impact from automated cars. Open can help in other ways too.

The data collected by cars is needed for them to do their job but automated cars will also use data provided by other things and people.

Automated cars won’t be like a starting character in the Civilization games. They’ll be able to see the full map. Civilization made by Firaxis Games, image from VentureBeat

An automated car will not wake up in a factory, blearily blink its headlights and then discover the world like a video game player constantly surprised by new things. The car will have a reasonably accurate map of the world, will get weather data (what sensible car would choose to drive into a hailstorm that might damage its paintwork?) and be able to share data with other cars.

Just as we hear of traffic jams from other people via radio alerts or smartphone apps like Waze, the people designing and building automated cars have planned for them to be able to share news about traffic congestion or improvements to their basic maps. Those improvements are vital because map data, just like any other data, is not always 100% accurate. Things change. An automated Google car driving down a street might discover that a road is blocked off, by sharing this with other Google cars it can make Google’s service more efficient.

This all sounds like good use of data, but it’s not good enough. We can and should do better.

Data should be as open as possible while respecting privacy

Werner Herzog’s first automated car looked a lot like a boat.

Sebastian Thrun of Stanford says in Werner Herzog’s new documentary “Whenever a self-driving car makes a mistake, automatically all the other cars know about it, including future unborn cars.”. But isn’t the only way that all self driving cars will ‘automatically’ know of all other mistakes is if the data that describes those mistakes is available beyond just the automated cars of one manufacturer?

At the Open Data Institute we think that we get the most value from data when it as open as possible while respecting privacy.

The team at OpenStreetMap’s 2016 April Fool’s spoof was a plan to launch their own automated car. They said: “our self-driving car breaks new ground by automatically correcting OpenStreetMap data based on your driving behaviour”. The story was a spoof but this bit — regardless of whether it’s OpenStreetMap or another mapping organisation/community relevant to a particular country or city — is one of the ways that cars can share data with each other and with other people.

Mapping is a shared problem. All cars, automated or not, will benefit from better maps. As will pedestrians, cyclists, local authorities planning new infrastructure investments, etcetera. Collaboratively maintaining open mapping data between all of these people can reduce costs and improve quality. Facebook are happy to collaboratively maintain open mapping data as they recognise the value in this approach. Automated car manufacturers, mapping organisations and policymakers should be too.

Reducing accidents is another shared problem. The machine learning algorithms that will drive automated cars will learn faster and more accurately from more data. Sharing detailed data about the conditions in place when an accident occurred will save lives.

People will ask an automated car to drop them off at an address. That address may not be in the current list of addresses — perhaps it’s a new flat? — so the person may teach the automated car where it is. The address could then be sent to an open address register as a potential improvement to the data. The next automated car will know about it but addresses are vital for many other things from pizza delivery to an ambulance. We should be maintaining addresses as efficiently and openly as possible. Collaborative maintenance helps with that and openness means that anyone can use it.

There will be many other types of data collected by the car that when opened up in this way will improve transport services, save lives and make things better in other sectors.

Live weather conditions (something that the lovely folk at TransportAPI are working on). Air quality. Congestion data. Aggregated movement of people around a city. Etcetera.

This impact of opening up this data will be felt not just in better automated car services but in other services and sectors that use the same datasets. Automated car manufacturers are in the transport business, not the mapping or air quality business. Publishing the data openly will help them tackle shared problems and increase the impact of the data. Everyone benefits from better and more open data.

Automated car data should be secure, private and open by design

The Open Data Institute’s data spectrum. The most important things about data is who can access and use it. Mapping an automated car’s data against this data spectrum would be very interesting.

The transport sector has long been a leader in open data. The countries and organisations that have taken the lead in opening up this data have benefitted both from better services for people and through the creation of innovative new services like GoogleMaps and companies like CityMapper, Transport API and ITOWorld that create jobs and help get the data used.

As that seemingly inevitable next wave of change occurs with the rollout of automated cars that will improve transport, free up space, save lives by reducing accidents and impact jobs let’s make sure we don’t forget about the data infrastructure that is necessary for those cars do their job and can create so much value for the rest of our society.

Making that data infrastructure secure, private and open by design will benefit everyone.

If you want to chat about the thoughts in this blog then tweet or mail me.