Are we ready for driverless vehicles on Australian roads? In my January 2019 blog post I predicted driverless vehicles as one of the top 4 technology trends for 2019, but are we there yet?
I spent the last week of October in Sydney, to attend the International Driverless Vehicles Summit at Sydney Olympic Park and the CEBIT Australia annual technology conference. But sadly nothing I learnt at the technology events in Sydney convinced me that we’re ready for driverless vehicles on Australian roads. Driverless trains, buses and cars are certainly the way of the future, but we must have the supporting transport infrastructure in place to bring driverless vehicles into mainstream use.
The most difficult driverless technology to bring into common use is driverless cars. Whilst many cars today have a fabulous array or assistive technology, such as emergency braking and lane keeping assist, we’re a long way from having fully autonomous cars trusted on our roads.
At the Driverless Vehicle Summit I listened to Michael Milford from the Queensland University of Technology talking about their tests of off-the-shelf software used in driverless cars.
They found that the combination of cameras and LIDAR with off-the-shelf software was not enough for the cars to drive themselves on our roads. The cars couldn’t recognise street signs and traffic lights in poor visibility, such as in the dark and the rain. There were lots of false positives, for example the car seeing a speed limit sign on an adjacent slip road alongside a freeway.
To solve the problem of signage confusion driverless cars will need to operate in collaboration with smart infrastructure, such as digital signs and traffic lights that can communicate to the cars.
Recognition of lane markings was another cause for concern, with the reflection off water sitting in a rut along the road and white paint spilled in a line along the road being confused for lane marking.
The most difficult problem to solve before we can trust driverless cars is that they don’t know what they don’t know. It’s difficult for the cars operating autonomously to know when they’re not understanding the environment correctly, and that is when things start to go wrong.
Improved results were gained by combining the off-the-shelf technology installed in the cars with the use of maps of the streets the cars were being tested on. Once the driverless car combines the inputs from cameras and LIDAR with its known map of the environment, it can operate more consistently and accurately. Creating a map of all Australian roads is going to be a huge undertaking. But with multiple cars providing data to create a combined map, this will enable a commonly understood map to be used with confidence for driverless cars.
If driverless cars are going to share mapping data, they could also share live-feed data about the environment, with all driverless vehicles and the traffic infrastructure sharing information about hazards and changing driving conditions.
For driverless cars to be brought into common use there needs to be a combination of technologies – in-car software, vehicle cameras and LIDAR, GPS positioning and maps, smart traffic infrastructure providing data to the cars and shared communication between driverless vehicles about their environment and driving conditions.
One of the biggest risks on our road is driver fatigue and inattention. Cars are now being released to the market with assistive technology which detects driver fatigue by cameras constantly scanning the driver’s face for signs of sleepiness. In the future, driverless cars are expected to solve the problem of driver fatigue and inattention.
Another speaker at the Driverless Vehicle Summit was Angus McKerral from the University of Newcastle, talking about their research into attention and driver fatigue during automated driving. They tested drivers in a simulator to measure their fatigue and inattention levels after 45 minutes actively driving versus passively sitting behind the wheel of a driverless car.
The finding of the study was surprising. Drivers were more fatigued by passively sitting in the car than by actively participating in controlling the car. This is a serious problem, as the technology of cars which can drive in autonomous mode still needs a human to intervene when they get lost in the environment, in an emergency, or when the autonomous mode fails. And current legislation requires active supervision of cars which are being driven in fully autonomous mode.
The University of Newcastle team re-tested the drivers using the simulators, this time allowing the drivers behind the wheel of cars in autonomous mode to do non-driving related tasks, such as reading on a tablet device. This was found to increase driver attention and reduce fatigue, as it gave the drivers something to keep them alert rather than just sitting passively doing nothing. But these non-driving related tasks are a distraction from what’s going on outside the car and of course not legal under our current traffic laws.
I believe the main problem people have with driverless technology is giving up the illusion of control. With a human behind the wheel, it’s up to us to decide what we should do when things go wrong. But ceding this decision making to a driverless car’s algorithms is problematic. For example, there’s debate about whether the technology should sacrifice the life of the person in the car in the event of an unavoidable collision. Should the person in the car be sacrificed or should the car or a crowd of school children in the path of an accident be sacrificed?
The difficult transition period while we deal with these issues is similar to what society faced when horse-drawn transport was being replaced by motor vehicles. When the technology was first introduced in Britain a 4 mph speed limit (2 mph in town) was legislated for road vehicles, there had to be at least three people be on board, and somebody waving a red flag was required to walk in front of the car at all times. Meeting an oncoming horse meant the road vehicle had to be brought to a full stop!
I was pleased to note that the two driverless shuttle buses on display at the Driverless Vehicle Summit both have connections to businesses operating in Adelaide, Easymile and SAGE Automation. The Ollie and Matilda driverless bus and pod (which I mentioned in my October 2019 blog post on technology in tourism) were being demonstrated at the Driverless Vehicle Summit.
The other bus being demonstrated was the Navya shuttle trial, which runs a small driverless bus around the public roads at Sydney Olympic Park. It’s a small step towards putting driverless buses into more common use but relies on smart infrastructure (traffic lights and smart bus stops) at the Sydney Olympic Park precinct to communicate with the bus.
Transport NSW have announced they intend to run full size driverless buses on their busiest routes by 2022. But this will only be feasible once smart traffic infrastructure is in place to support the buses in operation and the buses have been trained with maps of the streets their bus routes follow.
While in Sydney I also took the time to experience a ride on the Metro, Sydney’s new driverless train that runs from Chatswood out into the north-western suburbs. The automated doors work in perfect harmony with the glass doors along the platform when the train pulls into each station and it was fun to stand at the front of the train, where the driver’s cab would normally be, and watch the track whizzing past.
This driverless train is an ideal use case for driverless vehicles as it runs in a closed system, once it leaves Chatswood railway station it heads down into a tunnel and it speeds along happily in isolation from any possible interference from pedestrian or vehicle traffic.
However, converting the remainder of our train networks into driverless vehicles will require a huge investment in infrastructure to ensure the trains can be trusted to run autonomously.
Driverless cars are still very much in test mode, the technology’s not yet up to releasing them in the wild.
Driverless buses are becoming increasingly possible, with a number of shuttle buses running in controlled environments, such as the Navya bus at Sydney Olympic Park and the shuttle bus which runs around the Tonsley innovation precinct in Adelaide.
But the best use case for driverless vehicles is trains. Which makes me, as a self-confessed train geek, very happy indeed.
Sydney Metro in meltdown after 'mechanical problems' on train https://www.smh.com.au/national/nsw/sydney-metro-meltdown-after-mechanical-problems-on-train-20191031-p5360q.html
Sydney’s driverless shuttle bus service at Olympic Park https://www.itnews.com.au/news/sydney-driverless-shuttle-merges-into-live-traffic-533046?eid=3&edate=20191028&utm_source=20191028_PM&utm_medium=newsletter&utm_campaign=daily_newsletter
Driverless buses trialled on Sydney's busiest routes by 2022 https://www.itnews.com.au/news/driverless-buses-trialled-on-sydneys-busiest-routes-by-2022-518815