Automation of transport: Hail technology

Posted Mar 3, 2018 by Tim Sandle
A new study shows how deep learning can be used to help predict when people need rides. So-called hailing technology, signalling further automation of travel, could further disrupt transport-as-a-service.
Thousands of taxi drivers gather in Bogota  on March 14  2016 to protest against the Uber taxi-booki...
Thousands of taxi drivers gather in Bogota, on March 14, 2016 to protest against the Uber taxi-booking mobile service, leading to traffic congestion across the Colombian capital
Guillermo Legaria, AFP
Penn State university technologists have carried out a study which shows computers are probably better at predicting taxi and ride sharing service demand than manual systems. This could also pave the way towards smarter, safer and more sustainable cities, fitting in with smart city innovations.
For the research, the technologists looked at two types of neural networks (which are computational systems modeled on the human brain). These networks collected and studied patterns of taxi demand. By deploying deep learning approach the combined networks were able to predict, with a high degree of reliability, the demand patterns significantly for taxi services. The predictions were higher than current technology and far better than manual systems requiring human intervention.
For the study, the researchers drew upon data for taxi requests made in Guangzhou, China, from Feb. 1 to March 26, 2017. During this period there were around 300,000 ride requests each day. With the historical data, the neural networks assessed time and location of requests. Based on this analysis a computer was able predict how the demand changes d over time.
This technology would not only assist taxi companies with predicting demand and with better scheduling, there is also an environmental benefit in terms of reducing the number of fleet cars on the roads during periods of low demand. Furthermore, the technology could lessen the time that taxis stand idle waiting for rides (where the engines are often running). To add to the environmental credentials, safety can be included. Since accidents often occur in congested areas, improved ride prediction technology can help to reduce accident patterns during rush hours.
According to lead researcher Professor Jessie Li: "You can imagine how important it would be to predict the taxi demand because the taxi company could dispatch the cars even before the need arises."
The research was recently presented at the AAAI Conference on Artificial Intelligence. The conference took place in New Orleans, Louisiana, U.S., during February 2018.