Smarter Urban Dynamics - Carbotraf


Carbotraf

The CARBOTRAF is an 7th EU Framework Programme project which aims to realize a method, system and tools for adaptively influencing traffic in real-time to reduce carbon dioxide CO2 and black carbon (BC) emissions caused by road transport in urban and inter-urban areas.

The CARBOTRAF consortium brings together 8 European expert institutions from 3 EU Members States. It includes research organisations and universities (AIT, VITO, IMPERIAL COLLEGE LONDON, OSTERR. FORSCHUNGS- UND PRUFZENTRUM ARSENAL) and industrial companies (IBM, AIR MONITORS, EUROPEAN TECH SERV., EBE SOLUTIONS).

One of the major challenges city operators are facing is to make decisions based on constantly changing traffic conditions to meet defined objectives, such as reducing CO2 and Black Carbon emissions and reducing traffic related congestion. Importantly, operators need to be informed about unusual traffic conditions and their potential impact while they are emerging as opposed to working out strategies in a reactive manner. These requirements puts a great emphasis on an IT infrastructure that is capable of analysing vast amounts of data in real-time and delivering actionable results in a timely fashion. As part of the EU FP7 project CARBOTRAF, IBM Research - Ireland is advancing the state of the art in traffic analytics by developing a scalable streaming toolkit utilising IBM's InfoSphere Streams platform (version 2.0.4). This Intelligent Transportation Analytics toolkit is capable of making sense of geographically referenced sensor data from diverse and heterogeneous data sources (e.g., SCATS - Sydney Coordinated Adaptive Traffic System). It integrates point estimates of traffic volume at the intersections into traffic densities along the street segments in real-time using a macroscopic model adhering to the conservation law of flows. This approach also provides insights into urban-scale travel times and queue lengths building up at the intersections. In CARBOTRAF these real-time traffic state estimates are used to map traffic conditions to emission profiles given a traffic operators unit of control, i.e., traffic actions such as traffic signal control, imposing speed limits, and variable message signs informing travellers about expected delays along certain routes. This provides a "What-If" analysis of the current traffic state and all available traffic actions.

Given historic data, the key performance indicators (congestion, CO2/BC emissions, etc) along street segments can be computed using microscopic traffic flow simulation and used to train a decision support model assisting a traffic operators in making better and faster decisions before congestion is at risk of spilling over into upstream street segments and thus blocking traffic at the intersections. Using a large repository of historic data, IBM Research - Ireland is using IBM SPSS to train decision trees that capture the impact of all traffic actions given the current traffic state and estimates of related predictions on the key performance indicators. These decision trees are integrated into the IBM InfoSphere Streams Intelligent Traffic Analytics workflow to assess in real-time the impact of certain traffic actions. As a consequence, the operator is informed about the estimated effect and uncertainties of optimal traffic actions on meeting desired objectives. This is in stark contrast to how traffic operators currently work. Typically, a control room is equipped with monitors to observe CCTV footage at critical intersection as well as relying on travellers submitting information through radio stations. An inherently data-driven approach, such as the one envisioned as part of the CARBOTRAF project has the potential of cutting down on the time it takes until operators know of incidences or other developments that require immediate action.

For more information, see the project website.