British drivers for ride-hailing company Uber are taking their employer to court in the Netherlands in order to access secret performance data the company collects on them.
- DriverML: a machine learning algorithm for identifying driver genes in cancer sequencing studies Nucleic Acids Res. 2019 May 7;47(8):e45. Doi: 10.1093/nar/gkz096.
- Driver's License Calculator: Florida. Calculate your Florida Driver's License number from your information. Reverse analyze an existing number.
- Driver drowsiness contributes to many car crashes and fatalities in the United States. Machine learning algorithms have shown to help in detecting driver drowsiness. We try different machine learning algorithms on a dataset collected by the NADS-1 1 simulator to detect driver drowsiness.
- Identifying driver mutations in a patient’s tumor cells is a central task in the era of precision cancer medicine. Over the decade, many computational algorithms have been developed to predict the effects of missense single-nucleotide variants, and they are frequently employed to prioritize mutation candidates.
British Uber drivers take company to court over secret algorithms. The drivers claim Uber uses a secret algorithm to allocate rides and fares based on data it collects about drivers' performance.
According to the legal challenge the drivers say Uber determines how to allocate rides and fares based on information it generates about drivers' performance, behaviour and other personal traits it infers through the app they install on their phones.
They are demanding full access to this information and to details on how Uber's algorithms work, claiming that the it ultimately decides how much money they can earn.
'Uber should offer total transparency,' lawyer Anton Ekker told Reuters at the District Court in Amsterdam, where Uber's international head office is located.
In response, Uber said it had shared all the information it was required to with its drivers, but that sharing any more could impact passengers' privacy.
Please use Chrome browser for a more accessible video player
'It's extremely dangerous for the privacy of passengers if we share all data about specific rides,' said Uber's lawyer Axel Arnbak said.
'This is very sensitive information. For a driver, it simply says they went from A to B, but it tells much more about a passenger,' Mr Arnbak added.
More from Netherlands
The company said that rides are primarily allocated based on driver locations, routes and fare preferences - and that drivers were excluded from giving rides to passengers who had previously given them the lowest possible rating.
However, it added that individual ratings, complaints and other feedback could not be shared with the drivers for privacy reasons.
The District Court in Amsterdam said it would make a decision on the claim by 11 Februrary 2021.
Turn right, turn left, turn right: inside Orion, the 10-year effort to squeeze every penny from delivery routes
TIMONIUM, Md.—Here’s a math problem for you. Each United Parcel Service Inc. driver makes an average of 120 stops per day. There are 6,689,502,913,449,135,000,000,000,000,000,000,000,000,000, 000,000,000,000,000,000,000, 000,000,000,000,000,000,000,000,000,000,000,000,000, 000,000,000,000,000,000, 000,000,000,000,000,000,000, 000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000 alternatives for ordering those stops. Which option is the most efficient, after considering variables such as special delivery times, road regulations, and the existence of private roads that don’t appear on a map?
Even if an optimal answer exists, the human mind will never figure it out. And while experts at UPS have been giving the problem their best shot for more than a century, the company is shifting that work over to a computer platform called Orion, which is 10 years and an estimated hundreds of millions of dollars in the making. “Can a human really think of the best way to deliver 120 stops? This is where the algorithm will come in. It will explore paths of doing things you would not, because there are just too many combinations,” says Jack Levis , senior director of process management at UPS.
The 1,000 page Orion algorithm is an exercise in heuristics, written by a team of 50 UPS engineers in Timonium, Md. Instead of searching for the optimal, or best possible answer, heuristics is the search for the best answer one can find, the results continually refined over time, based on experience. Orion consists of many components, including a “traveling salesman” algorithm, a familiar tool that calculates the most efficient path between a variety of points, and geographic mapping. What makes Orion unique is the way it puts these elements together, striving for a balance between an optimum result and consistency, according to Mr. Levis. “Customers and drivers like consistency. Orion has to know when to give up a penny to make the results more stable,” Mr. Levis said.
None of the solutions that Orion spews out are big or dramatic. It is all about saving a dollar or two here and there. But in a network with 55,000 routes in the U.S. alone, that adds up. “In our business, small things mean a lot. If you can reengineer process, the gains will be greater than you think,” Mr. Levis said.
Such savings matter to UPS, which is struggling with a tighter-margin business and a union workforce that is compensated at the high end of the industry scale. Its challenges are unique. Rival FedEx Corp. uses an independent contractor model for its ground network, so it’s not ultimately responsible for miles driven to most of its residential stops.
A Changing Business
E-commerce has shifted more and more of UPS’s delivery stops to residences, and those packages are expected to make up half of all deliveries by 2018. It’s a radical change from 15 years ago, when drivers would drop off several packages at a retailer. Now, they make scattered stops to drop off one package at houses in a neighborhood, driving further and taking up more time.
On Nov. 13, UPS CEO David Abney said he expected Orion to save the company $300 million to $400 million a year, once it is fully implemented in 2017. The more than 40% of the company’s 55,000 U.S. routes already using the software at that time had been reduced by an average of between seven and eight miles, the company said. The company can save $50 million a year by reducing by one mile the average aggregated daily travel of its drivers. Those savings are critical as UPS tries to boost earnings growth, which has been in the 5% range in recent years and dipped in 2014, as low-margin deliveries related to e-commerce become more prevalent and the company scrambles to figure out how to manage its holiday season.
While Mr. Abney cautioned that at least some of Orion’s gains would be offset by rising costs related to delivery of its customers’ e-commerce orders, he is targeting per-share earnings growth of more than 50% over the next five years. The company lowered its 2015 outlook earlier this month.
UPS won’t say how much money it has invested in Orion. But management and information technology expert Thomas H. Davenport, a distinguished professor at Babson College near Boston,believes Orion is the largest deployment of operations research, and that UPS spent $200 to $300 million to develop it, excluding many years of investments in underlying driver technology and communications infrastructure.
How Orion works
A driver—in this case, let’s use the example of Tim Ahn, who has been a full-time driver for 20 years, currently with a route in Gettsyburg, Pa.—would use his UPS tablet, known within the company as a delivery information acquisition device, or DIAD, to punch in at the beginning of his shift, as he does now. The DIAD would show him two possible ways to make his deliveries, one using Orion, and one using the current combination of work rules, procedures and analytic tools that are used to establish the order of package deliveries. He can choose to work in either way, but if he decides not to use Orion, he will be asked to explain the decision.
Algorithm Driver Updater
ENLARGE
Orion already has been at work for hours, though. It may have reordered Mr. Ahn’s schedule of stops for the day hundreds of times, as packages were added to the list assembled before he arrives at work, and as customers used the company’s My Choice self-service platform to change the time or location of their deliveries. UPS says My Choice membership has grown steadily since its launch in 2011 to 12.9 million today.
At one point, Mr. Ahn was scheduled to start his route at 8:45 a.m., making 125 deliveries and traveling 117.85 miles during the day. But now Customer 1 wants a package delivered between 11 a.m. and 1 p.m. That stop was originally scheduled by Orion for 1:25, so Orion has to recalculate. It considers up to 200,000 of the best options before settling on one. The package will now be delivered by 12:30 p.m., adding 1.39 miles to the day’s route, at a cost to UPS of $1.99. It takes Orion and the network about eight seconds to return an answer.
Now, Customer 2 specifies that a package that Orion originally scheduled for delivery by UPS at 3:51 p.m. must take place between 4:30 p.m. and 6:30 p.m. Orion considers a range of options before settling on a delivery order that arranges the delivery for 4:46 p.m., adding 1.64 additional miles and $2.77 in cost.
Orion is a useful tool, according to Mr. Ahn. “Orion had me do things in the morning I would not think of doing, and it saved me miles later in the day,” he said.
Driverless Algorithm
Rough Patches
The deployment of Orion isn’t always so smooth, though. That is where Mr. Levis comes in. As project manager, he is responsible for getting people and machines to work together. During the earlier stages of writing the Orion algorithm, it was Orion that had to learn to accommodate people.
ENLARGE Jack Levis is the leader of UPS’s Orion project, which has been some 10 years in the making. Photo: Matt Roth for The Wall Street Journal
Algorithm Driven
“The project was nearly killed in 2007, because it kept spitting out answers that we couldn’t implement,” Mr. Levis recalls. The earliest versions of Orion focused on getting the best mathematical results, with insufficient regard for the interests of the driver or the customer, who value some level of routine. For example, regular business customers who receive packages on a daily basis don’t want UPS to show up at 10 a.m. one day, and 5 p.m. the next. And a customer who is expecting a shipment of frozen food needs delivery as soon as possible, even if efficiency demands that someone gets priority.
To get the project back on track, UPS chief scientist Ranga Nuggehalli turned to Bob Santilli, a senior project manager, asking him to describe a perfect route. Several weeks later, Mr. Santilli came back with the results of his effort, which produced a model plan of stops for drivers on a route in Lancaster, Pa. The engineering team extracted proprietary rules from the Santilli route and built them into Orion.
“By April or May of 2007, he had the first working version of Orion, which balanced consistency and optimality. It had to do with keeping the driver in a path. The route should flow. That is what we learned. That is what brings consistency. Orion can make exceptions to the flow, but it has to do so in an intelligent manner and it can’t make an unlimited number of exceptions,” Mr. Levis said.
55,000 Routes
The process of balancing Orion’s logic with the real-world experience of drivers is built into the rollout of the project. A team of 700 trainers is working its way through all 55,000 U.S. routes, deploying Orion to one UPS facility at a time, a process expected to be more than 70% complete by the end of the year.
Algorithm Driver's License
It takes about six days to train a driver. The first day of training is spent fixing maps, as the trainers pour over satellite images and talk to drivers about minute details of their routes. On the third day, the trainers ride the route themselves in a rental car. On the fourth and fifth day, the trainers ride with the driver, and try to figure out what Orion is getting wrong about the route. More revisions are made on the fifth day, and a final ride-along occurs on day six.