Truck fleets led the way in commercial adoption of mobile communications for tracking and navigation, and now 30 years later they are repeating the trick for autonomous driving. Trials and even early pilot deployments are well underway, with Uber and Google sister company Waymo both announcing deployments just a month after settling their strange case over alleged theft of intellectual property.
That case unearthed more of Uber’s dirty secrets but failed to establish any clear evidence of real IP theft associated with an ex-Waymo employee who subsequently joined Uber. It was settled by Uber paying $245m purely in equity with no legal fees either way, apparently as compensation for not having gone by the book, rather than being guilty of any IP appropriation. The two are now more partners than adversaries given Waymo’s 0.3% equity stake in Uber – and mutual interest in this industry’s progress.
Both are pushing autonomous trucks hard in the belief that this will yield ROI before passenger cars, as a proving ground for the technology that will then be rolled out in conventional cars. That emphasis may be intensified further for Uber after it suspended self-driving car tests following the death of a pedestrian in Arizona.
There are several reasons why trucks are leading the autonomous drive. They tend to follow main roads, sometimes in convoys, traveling between out of town depots or factories without entering urban areas or narrow streets. They also tend to take long journeys where driver fatigue is an issue, so that even partial autonomous driving could provide relief and cut costs by increasing daily journey times, subject to relaxation in regulations over driving hours and rest periods. This could lead to fuel savings by spreading out journeys over longer periods at lower speeds.
Uber has tried to play the good citizen here, by arguing that, for the immediate future at least, its robot trucks will not cost jobs since the vehicles are not yet ready for full runs even between loading docks. They will require ‘safety’ drivers behind the wheel at all times, ready to take control in an emergency or perhaps in heavy traffic. In fact, more jobs may be created because the system will operate a bit like freight rail.
Uber Freight will operate conventional trucks to pick goods up from customer premises and deliver them to a hub outside town. They would then be transferred to a robotic truck with safety driver on board which will transport them to a destination hub and then quite possibly via another conventional truck to the final point of delivery. This could mean three drivers are involved where one would have been before, although with much shorter journeys at each end the gain would not be as much as that.
The company has been coy about details, save to claim that its robot trucks have so far collectively driven 2m miles in trials. The obvious next step is to eliminate the human-driven short hauls at each end and then net employment might start falling. It will be a long time though before technology or legislation will allow elimination of the safety driver.
With its pedigree as a taxi booking app, Uber naturally looked first at autonomous driving for cars with a long term view of cutting costs further by eliminating drivers, which would also increase passenger capacity. But then it saw trucking was going to be a bigger early market for progressive levels of autonomy and so purchased long haul trucking start-up venture Otto for around $650m mid-2016.
It was this acquisition that immediately kindled the bitter legal dispute with Waymo, which alleged that Otto was merely a shell company set up as a receptacle for technology stolen by Anthony Levandowski, the ex-Google engineer who founded the start-up.
Even Otto does not manufacture the vehicles itself but adapts trucks made by Volvo (with technology quite similar to that used by Waymo and Nissan as it happens), including radar, cameras around the vehicle body and also of course Lidar (light imaging, detection, and ranging) for detecting distances to objects by measuring the spot time taken for laser pulses to reflect back to the sensor. Otto started on Volvo 780 semis by adding self-driving technology in August 2016, testing them on US interstate highways which are often straight and relatively uncongested.
With these trials going reasonably well Uber then decided it needed a business framework for operating autonomous trucks and so bought the five-person team behind a Chicago transportation brokerage from 4Front Logistics. This brought in expertise in connecting manufacturers or retailers that want to ship goods with truck owners and fleets offering the transportation. In this highly fragmented US brokerage market, Uber saw an opportunity to gain some competitive leverage.
Meanwhile, Waymo has just launched a pilot program in Atlanta focusing just on self-driving trucks and automated logistics in partnership with sister company Google as a guinea pig as well as provider of various technology elements and additional self-driving expertise.
The choice of Atlanta for the Waymo truck trial was no accident as Google has a number of data centers in the surrounding region and the vehicles will be tried out on deliveries to them. This arrangement will allow other aspects of the delivery ecosystem to be evaluated with a view towards greater levels of autonomy, including allocation of loads, while connecting shippers, factories, distribution centers and ports into the existing freight shipping system.
Google says it makes great use of simulation to develop capability for dealing with exceptions that arise just rarely, similar to the way human airline pilots are trained. The figure cited of fivebn miles driven in simulation is somewhat meaningless but does emphasize the importance of that in autonomous system development. Even so there is no substitute for testing on the road even though that brings risks of real accidents and Waymo says it has so far accumulated 5m miles there.
Also like Uber and others, Waymo is not waiting on governments to approve new vehicle exemptions that would allow it to deploy futuristic autonomous vehicles devoid of steering wheels or pedals, such as General Motors’ proposed robo-chariot planned for launch in 2019. Much better, they argue, to work on modified versions of existing vehicles because these are what people will be using in trials and early deployments for some years to come until regulators, who are naturally cautious, see fit to allow the steering wheel to disappear to usher in full SAE Level 5 autonomous completely devoid of human intervention.