Silicon Valley start-up demonstrates its self-driving technology in sketchy conditions
Can an autonomous, driverless vehicle drive safely through busy city traffic on a dark, rainy night? You bet it can, if this new video from Silicon Valley-based start-up Drive.ai is any indication.
Indeed, as you can see from this 3:46 time-lapse film shot in a car as it navigates the dim, wet urban streets of Mountain View, California with no human intervention whatsoever, it can even handle a heavy downpour, narrow streets littered with parked cars, glare from oncoming traffic, a busy four-way stop sign, a car possibly illegally cutting in front of it, a broken red light and confusing reflections from a dark, wet road.
Drive.ai says it’s devoted to developing the software of self-driving vehicles to create a safer, more productive future on our roads. Most of its staff have made their way from Stanford University’s Artificial Intelligence (AI) lab and the company is one of around 20 now with a license to drive autonomously in the state of California.
Carol Reiley, co-founder and president of Drive.ai, explained in an interview with TechCrunch how deep the team is going with the artificial part of the intelligence algorithms in autonomous vehicle software.
“We are using deep learning for more of an end-to-end approach. We’re using it not just for object detection but for making decisions, and for really asking the question ‘Is this safe or not given this sensor input on the road?,” she explained.
“A rule-based approach for something like a human on a bicycle will probably break if you see different scenarios or different viewpoints. We think that deep learning is the definitely the key to driving because there are so many different edge cases.”
The Society of Automotive Engineers (SAE) uses a scale of six levels to define different levels of automation. Those in the field of building AI for cars have adopted both the definitions and the challenges they entail.
For example, current driver-assist systems like radar-based automatic cruise control and camera-based lane-keeping — which can accelerate, stop and steer a car without driver input for short periods — are classed as Level 1 autonomy, while Tesla’s fully autonomous – and controversial — Autopilot system is rated as Level 2.
A wide range of car-makers – as well as the likes of Uber, Google and Apple — are aiming to deliver Level 4 production cars by early next decade. Drive.ai is already showcasing this technology and is already working on Level 5, in which there should be no need for a ‘driver’ at all.
Tesla claims its vehicles are already fitted with the hardware to support Level 5 autonomy, but hands-free driving remains illegal in most jurisdictions, including all Australian states.
Since human error is the cause of 90 per cent of collisions, autonomous vehicles have been hailed as the cure for road deaths and have even been predicted to kill the car insurance industry, while Nissan promises that even car enthusiast will love self-driving cars.
Car-makers are investing billions in automotive AI and a driverless vehicles have already been approved for roads in the US, Europe, UK and Australia, where the government commenced autonomous motoring hearings this week and Victoria’s first self-driving vehicle trials start on Melbourne’s CityLink within months.
But despite assurances from companies like Audi and Volvo, which have vowed to take responsibility for any damage, injuries and deaths caused by their autonomous vehicles, most Americans doubt they’ll be more productive in an autonomous cars, there are fears of an increased risk of cyber attack from hackers and many remain sceptical of claims that self-driving vehicles will reduce traffic congestion.
Furthermore, going beyond Level 5 autonomy requires not just vehicle-to-vehicle (V2V) communication – which would require all vehicles to be fitted with a transponder — but expensive vehicle-to-infrastructure (V2X), as well as a high level of AI, raising serious ethical issues like ‘who should be killed if a fatal collision is unavoidable?’ and ‘will your car decide to kill you?’.
What this video doesn’t show is Drive.ai’s other areas of interest — tracking pedestrians and giving vehicles the ability to emote their intentions to pedestrians and other drivers — but it does demonstrate that groups like Drive.ai are hard at work to make the transition to a driverless future as seamless as possible.
Now let’s see the Drive.ai crew survive the swagger of an oncoming B-triple on the Nullarbor, and emerge unscathed from the gibbers, washouts and errant ‘roos of the Birdsville track.