Video feeds, country specific traffic sources for NL, UK
Elements define a location in time and space
Dataobjects linked to elements
Industry standard techniques
Standardization of data sources as a first step
Link, nodes, segments, areas
Algorithms work wit standardised data
Queue tracking on motorways and cities
Rich html5 app for browser and mobile
Interactive map with options to dinamically adjust time
Situational Awareness: focus only on extra-ordinary eventualities
Traffic 360, zooming-in/out adjusts datasource volume and intensity to reflect scope
Leveraging NodeJs, LeafletJs, Meteor
Beside the regular static representation of spatial data, Dante includes by default the time dimension and presents dynamically periodic changes, without hindering the rendering process and the user experience.
Dante is a modular server platform from the ground up written in Java using the OSGI standard. Currently there are well over a hundred modules that can be combined in different ways to achieve a broad range of functionality. Below we list some of the frequent combinations.
Assimilate any data source
Any type of data source can be assimilated in Dante. So far the following sources have been made accessible: open streetmap, openLR, loop detectors, bluetooth, floating car data, radar detector, video, weather imagery, structured feeds e.g. RSS, Datex2, unstructured data from twitter or websites. In general connecting to an API, accessing a known database or assimilating data directly from a collection of sensors is trivial.
Standardization of data: become source agnostic
All traffic observations and meta data are standardised in the so called D-State where D Stands for discretisation. The DState has a short term buffer of a few minutes where all data from all connected asynchronous sources can be placed correctly. At any point in time the DState tells us what the state currently is and recently was. The DSTate keeps track of the original data but also aggregates these into usable components such as: speed, travel time, density, flow, occupancy, blockage.
Tracking congestion phenomena
On top of DState we build ‘Trackers’. Trackers are advanced algorithms that track a specific traffic phenomenon. They are not data source dependant anymore but operate on the DState.
Shockwaves: detection and ultra short term prediction. Shockwaves are stop-and-go traffic jams that move upstream through the traffic network. Shockwaves are sometimes known to induce congestion where traffic is at a critical stage.
Motorway congestion: ‘regular’ congestion associated with a temporal/spatial capacity problem. During the peak hours at a certain (well known) bottleneck congestion starts. Vehicles start queuing until the peak demand falls below a certain threshold and the queue starts to dissolve.
Urban queue: The queue of waiting vehicles before a traffic light or urban intersection which have to wait one or more cycles before they can pass the traffic light..
Gridlock: The congestive condition when an urban network as a whole becomes congested..
Accidents: A special case of congestion is related to accidents. On average about 25% of traffic can be associated with accidents.
All trackers come with two qualities. Phenomena can be tracked at the link level or at the lane level. The difference is the lateral detail which is higher when tracked specifically per lane.
Given the tracked phenomena over a longer range of time, we apply various algorithms to the data in order to gain more insight. We define ‘seed regions’ in the network where at a certain time and place the probability of regular congestion onset becomes significant. Similarly we can define the shock wave frequency at certain links in the network.
Given the tracked phenomenon, we enrich these by adding as much meta data as we can. The meta data considered may be some statistical properties, properties from structured sources like traffic operators or meta data automatically derived from unstructured data sources via twitter, or internet. The metadata itself is in turn used for different cases. It is widely regarded as very useful by traffic operators. And in our prediction engine it is used as a way to pre classify certain types of congestion before we start to predict them.
Based on the combination of tracked traffic phenomenon, the short time knowledge of the recent past in DState and the enriched meta data we employ various algorithms to make a prediction. They are based on statistics, traffic flow theory, exact/discrete analytics, Bayesian data assimilation and sometimes black box models.
Finally all information is made accessible via internet on our traffic browser. The traffic browser is accessible on all html5 capable devices and is build on standards such as leaflet, meteor and nodeJS. The traffic browser is simply put: a dynamic mapping application, just like Google maps but with dynamic data. The traffic browser offers highly 2way interactive user specific sessions where users can view the traffic state at any historic, recent or future time at any place. In the past 7 years we have learned how to visualize data in such a way it is kept understandable.