
JANE yawns and climbs the stairs from the subway at 145th Street, New York. Sheâs almost home. A stranger rises from a bench as she approaches, catching her eye. âJane Murphy?â She nods. âHereâs your package.â
This is the ultimate aim of a crowd-powered delivery system dreamed up by a group of Microsoft researchers. Fictional Jane never has to deviate from her normal route to pick up her package. Instead, it is sent via a chain of people â an algorithm calculates the fastest route using aggregated location data from New York tweeters. Eric Horvitz of , Washington, calls the concept TwedEx. The idea could make it possible to deliver purchases to customers on the move, as well as making it cheaper to send them.
Basic crowdsourced systems , which hire strangers from the internet to . But TwedEx is different because it taps into existing human journeys. All the sender need do is write the recipientâs unique identifier on the package, their Twitter handle, for example, and let the TwedEx algorithm and the crowd do the rest.
Advertisement
By learning peopleâs average movements from their past Twitter data, TwedEx predicts which people to hand a package to at intermediate locations based on the packageâs final destination. A user would tell the network they had a package, the system would work out the best route and then each person in the chain would be told who to give the parcel to, as well as where, and when.
ÂTwedEx predicts which people to hand a package to at each stage based on its final destinationÂ
Citizen couriers would be paid a small incentive to carry packages â depending how far out of their way they go to deliver or receive one.
In simulations, TwedEx works remarkably well. âWe see that the typical time to get across the country is just five hours,â Horvitz says, explaining that packages can make it from New York to San Francisco in that time even when they donât begin their journey next to the airport. Real world packages would be held up by factors such as airport security.
A real life TwedEx wouldnât need to track people in real time to set up a delivery chain. It would just send out messages to people along the possible route, asking them to make the exchange at a certain place and time. Bulky items could be divided up among multiple couriers.
TwedEx can cover 50 per cent of New York if each link in the delivery chain waits up to 30 minutes and deviates up to 100 metres off-course from their usual route at exchange points.
You can also expand TwedExâs coverage by increasing the amount of time an individual in the delivery chain waits at transfer points for the package to reach them, and the distance they must deviate from their normal path, Horvitz says.
A paper describing the system will be presented at the AAAI Conference on in Boston in July. So far, TwedEx only exists in a simulation, but Horwitz says Microsoft is discussing building an app for a real-world pilot.
Adam Sadilek, who worked on TwedEx but is now at Google, says the most viable initial scenario for TwedEx would be in poor countries. âYou can imagine using this for the distribution of vaccines,â he says.
Another group is already working on making this a reality. James McInerney of the University of Southampton, UK, told the in Boston earlier this month about a crowdsourced delivery system for rural communities in the Ivory Coast. McInerneyâs system maps peopleâs movements via their cellphone data logs, provided by telecoms firm Orange.
âThis would all be done by text,â says McInerney. âEach participant would receive a text if theyâve been chosen by the algorithm, with the pickup and destination information, and theyâd just have to follow their normal mobility patterns.â
In McInerneyâs model, delivery time to remote areas would be 28 days. This is much longer than a dedicated van would take, but he sees the network as a delivery route for regular supplies, rather than irregular purchases, as it would be in US cities.
At least one of the USâs largest companies is already thinking about using the crowd for deliveries. In March, that Wal-Mart is considering using its own customers to build a delivery network for goods bought online, and that the idea is in the early planning stages. A DHL concept called Bring Buddy was floated at Expo 2010 in China, but has since disappeared.
ÂWal-Mart is considering using its own customers to build a delivery network for goods bought onlineÂ
And Horvitz isnât the only one who dreams of postal addresses that match people, not places. co-founder Andreas Raptopoulos says the ultimate vision for his company is to build a delivery system that uses a network of drones (see âThe drone is flying it to you nowâ), with package delivery reaching the same level of abstraction as the delivery of email on smartphones. âAn email doesnât care where you are, we want to do the same with physical packages,â he says.
The drone is flying it to you now
Matternet is aptly named. The firmâs CEO Andreas Raptopoulos is building an internet for stuff, which can transport small items over short distances using flying drones.
The drones will be suitcase-sized and should cost less than $5000 each, while maintaining a 10-kilometre range. Each will have on-board sensing systems that are sophisticated enough to land the drone at drop-off targets that are smaller than GPS can handle.
âWe are in early phase discussions with the TNT courier company in Europe,â Raptopoulos says. âThey work with the UN as a food distribution partner and we are working out where Matternet fits in the infrastructure chain.â
This article appeared in print under the headline âYouâve got chain mailâ