Multiagents to break up traffic jams
DLR traffic researchers are looking for solutions to increasing freight traffic
Gernot Liedtke is working on making the future navigable – in the truest sense of the word. The department head at the German Aerospace Center (DLR) and his team are researching whether delivery vehicles in future could still carry out their duties in congested cities – and what solutions there might be. Will trucks soon be driving themselves?
It’s a burning topic. Growing volumes of online orders and dwindling warehousing of stocks in industry and commerce are leading to an increase in delivery traffic. More pollution is being pumped out and roads are more jammed than ever. Vehicle bans have already been put in place to reduce the fine particulate and nitrogen oxide emissions that make us sick, but all they seem to do is divert the problem onto side roads and harm the local trade. Long the exception, Germany is now considering introducing the city tolls that are already commonplace in other countries.
To be able to predict how measures like this will affect urban freight traffic, before any money is wasted, the DLR researchers have developed the multiagent model “Freight-TransportLab”. This can simulate the dispatch, transport and receipt of individual consignments and the corresponding driving routes.
“With our multiagent simulation, we artificially recreate the real world so that we can perform targeted experiments,” says Liedtke, who is also a professor of traffic research at the Technische Universität Berlin. This is not an application of artificial intelligence per se, characterised by self-learning algorithms. But real-world intelligence is being built into the traffic researchers’ simulations.
“There is no getting around a city toll,” Liedtke says. The simulations showed that the toll would best be levied “based on distance and time of day and across all networks.” It was also investigated whether freight forwarders would invest in new trucks and what types those could be. That is because the toll also depends on the type of truck. Furthermore, different consignments, such as furniture moving or small parcel deliveries, also require different types of truck, which would accordingly take different routes. The associated costs will necessarily influence the freight forwarder’s decisions.
Liedtke gives another example, of how package deliveries will develop in Berlin. The volume is currently increasing by seven to eleven percent per year. “On average, each Berliner receives one packet a week,” says the traffic researcher. Assuming this trend continues, one can make short to medium-term predictions of the resulting freight traffic. In this respect, Liedtke predicts a breakthrough in autonomous driving, albeit not in urban areas. “It is quite conceivable that drivers will get out of their 40-tonners before the Autobahn, and the trucks will continue driving overnight on their own.” At some point, there will even be cars without steering wheel, pedals or driver.
The team of 18 DLR traffic researchers started by conducting basic research with the multiagent systems, but also wants to put the results into practice, for example in cooperation with traffic planning offices, public companies and software developers. They are also investigating spin-offs. “It’s nice to be able to collaborate on developments that can influence the overall traffic, sometimes long-term, sometimes short-term,” enthuses Liedtke, who studied physics and earned his doctorate in economics.
By Paul Janositz for Adlershof Journal