Lyft releases data to speed arrival of ‘robotaxis’
In order to speed up the development of self-driving vehicles, Lyft is releasing “the largest public data set of its kind”. The ride-hailing specialist hopes this gives the industry a much-needed jolt. 'Robotaxis' could turn huge profits, but autonomous-vehicle research is in a slump.
Academics as well as Lyft's competitors will get access to over 55,000 3D frames of footage collected via camera, lidar and radar and hand-labelled by human reviewers, plus a drivable surface map and the corresponding high-definition spatial semantic data.
Lyft Level 5
This dataset is part of the overall data gathered over the past two years by Lyft Level 5, the company's autonomous-driving R&D division. Publishing it will help “level the playing field for all researchers interested in autonomous technology”, explains a post on Lyft's blog.
Or, as Luc Vincent, executive VP for Lyft Level 5, says: “We don't care about doing everything in secrecy; we care about accelerating the whole industry.”
Mr Vincent added that Lyft will continue to release additional data – and will host a competition to inspire use of the data, with $25,000 in cash prizes, a presentation and job opportunities, at the NeurIPS conference (8-14 December, Vancouver).
Lyft is not the first company involved self-drive R&D to publicise datasets of its own research. Automotive software specialist Aptiv – one of Lyft’s partners, together with Google’s sister company Waymo – has done something similar in the past.
One reason for Lyft’s move could be that the company’s R&D is not as advanced as that of its competitors, meaning it has less to lose by publishing (part of) its data. Lyft is currently working on its third-generation self-driving car, with new sensor array and a proprietary ultra-HDR-capable camera.
Another reason could be that it’s a ploy to attract talent – hence the competition, prizes and job interviews. Self-drive R&D is a highly specialised and very competitive field.
60% cost reduction
However fragmented the data and however self-interested its release, the fact that more and more data is becoming publicly available will undoubtedly help to boost research in a field that still holds great promise, even as the ultimate prize keeps receding into the future.
It’s estimated that eliminating the driver from the taxi business could drive down cost by as much as 60%. This could translate into both lower fares (for customers) and bigger profits (for operators like Lyft).
Level 5 autonomy is also likely to have major benefits for corporate fleets in terms of cost savings. But full autonomy is turning out harder to achieve than previously thought.
As recently as 2016, various industry experts and players were still predicting the arrival of Level-5 autonomy by 2020 or 2021. But as the task has proven more complex (and thus more costly), the goalposts have shifted. Additionally, self-drive research is in a slump since an Uber test vehicle hit and killed a pedestrian in Arizona last year.
The emphasis now is on intermediate goals: Level 5 autonomy, but only on highways first, or in geo-fenced urban areas, for example. And even that may be five to ten years away.
GM Cruise, General Motors's self-driving unit, this Wednesday announced it would delay the commercial deployment of its self-driving cars past its original target, the end of this year, as more testing was required.
General Motors's self-driving unit is working with Honda and GM itself to develop purpose-built autonomous vehicles. The company did not say when it now expects to deploy its self-driving ride-hailing service.
Lyft competitor Tesla – by way of its controversial CEO Elon Musk – is virtually alone in still predicting the imminent arrival of Level 5: by the end of this year. And by the end of 2020 (next year, mind you), Musk still aims to have one million robotaxis on the road.
That may prove as overly ambitious as Mr Musk’s mission to Mars…