As our streets grow more crowded, traffic jams aren’t likely to get more tolerable. The number of cars on the world’s roads is set to hit two billion by 2030 — double what it is today, and the number of traffic headaches will likely follow a similar trajectory. Navigation apps that promise to re-route drivers before heading into congestion could offer some help, but today’s versions are not up to the challenge. Existing solutions offer the same guidance for drivers going to destinations in a given region. Thus, these solutions tend to move congestion from one spot to another.
Together with a team at New Jersey Institute of Technology, I’ve developed two solutions to provide drivers with individual, proactive, real-time traffic guidance that avoids or mitigates congestion. The first solution is centralized and assumes a server that receives traffic reports from the cars, containing the current location, speed, and the origin and the destination of the trip, and it computes traffic guidance for each driver who could be impacted by congestion. The drivers then receive these individualized guidelines in real-time. This system can reduce the average travel time of the drivers by half in cases of severe congestion.
However, the system suffers from two problems. For one thing, generating new routes while staying up to date on real-time traffic maps requires significant computing and networking resources. User privacy is another growing concern: as more drivers communicate their location, origin and destination information to a central traffic-app coordinator, the more opportunities arise, unfortunately, to compromise that data.
More recently, together with my team, I’ve developed and tested a second solution that addresses these issues. We’re calling our approach DIVERT — because it’s a DIstributed VEhicular RerouTing system for congestion avoidance. The main innovation is seemingly simple but crucial: we turn a centralized system into a largely distributed one, shifting to individual “smart” vehicles the roles of communicating with one another and determining new routes. With this tweak, the problem of limits on server power falls by the wayside, and fears over privacy breaches recede in the rearview mirror.
The Problems with Central Coordinator-Run Traffic Apps
A centralized system for managing traffic routes, even for one major city, has a lot to deal with. Users expect real-time communication with their traffic apps, which means these systems need to constantly be sending updates on paths and receiving location updates from drivers. At the same time, intensive computation is required to crunch predictive algorithms and assign vehicles to new paths as necessary. This level of server activity renders centralized solutions unfeasible in large regions with many vehicles — which also happen to be the places where these apps are needed most.
Another problem that comes with reliance on a central coordinator to manage all driver information is that, well, that coordinator has access to all driver information. Not only do people behind the wheel have to frequently share their current location for these apps to work, but they have to share both origin and destination points for every trip. The idea of hackers gaining access to this information and matching it up with individual drivers is frightening, and not something anyone wants to expose themselves to.
The Solution: DIVERT
DIVERT is a hybrid system, meaning it draws from the benefits of both centralized and fully distributed architecture to create a superior system. Traffic apps must offer their users a global traffic view — which come from a central coordinator — but we can’t concentrate all information with that entity, for the reasons discussed above. Therefore, DIVERT does both: it uses a centralized server, reachable over the Internet, to offer an accurate global view of traffic, but it distributes computational load by making vehicles situated in the same region communicate with one another and locally compute new routes.
How it works: “smart”-enabled vehicles communicate with the server over a 3G or 4G network with reports of local traffic density and to receive updates on traffic conditions along the upcoming route. Vehicles that are closely located to one another share information directly to determine immediate traffic density, to pass around traffic data passed down from the server and to figure out a distributed re-routing strategy.
Offloading a significant portion of the re-routing computation in this way reduces the network load on the server by 95 percent, making these sorts of navigation systems much more scalable in the places where they’re most in demand. Most privacy concerns are automatically snuffed out, too: DIVERT never knows the origin-destination pairing of its users, for one thing. Our system also dramatically reduces the number of vehicle location updates sent to the server and only asks vehicles to upload location reports when they’re located in high-density areas (low-sensitivity areas, as far as privacy is concerned). Our hybrid system decreases the exposure of private information by an average of 92 percent.
So far, in terms of travel time, DIVERT’s performance is slightly lower than that of the centralized system, though the gains in privacy and scalability may account for the weaker performance in the minds of some drivers. Nevertheless, DIVERT still provides substantial savings in driving time when compared to no-rerouting option. As our world’s streets offer up more bottlenecks, jams and gridlock, traffic re-routing solutions will need to take into account computational load on servers and the public’s concerns around privacy in the light of increasingly frequent data breaches. DIVERT offers a promising glimpse of a route we can take to get there.
This article is published as part of the IDG Contributor Network. Want to Join?