ITSC-2022 What to Sense When There is no Sensor: Ex-novo Traffic Flow Estimation for Non-Sensed Roads
This GitHub is intended to gather all datasets, Python source code, and simulation results of the paper above.
A proper traffic characterization allows for operational measures towards handling congestion. However, traffic data must be collected beforehand for developing a forecasting model for a certain road location. Costs for traffic surveillance are driven by the time a sensor is deployed on the location to be monitored. A novel forecasting framework is presented, wherein a traffic flow prediction model is developed without the need for a constant input of traffic observations at the target location. Traffic estimations are computed by using data from permanently sensed roads, as well as a small sample of data captured in the target location. When to provisionally deploy a sensor on the target is formulated as a bi-objective optimization problem and solved by means of evolutionary meta-heuristics. The feasibility of this framework is assessed over a case study comprising real traffic data from Madrid (Spain). The obtained results show that traffic can be characterized by collecting a single week of data in the target location. Further experiments ensure that the optimization process can be trusted for selecting the schedule for provisional sensor deployment. The proposed framework paves the way towards cost-efficient traffic modeling over multiple non-sensed locations of a road network.
@inproceedings{manibardo2022sense, title={What to Sense When There is no Sensor: Ex-novo Traffic Flow Estimation for Non-Sensed Roads}, author={Manibardo, Eric L and La{~n}a, Ibai and Del Ser, Javier}, booktitle={2022 IEEE 25th International Conference on Intelligent Transportation Systems (ITSC)}, pages={1668--1675}, year={2022}, organization={IEEE} }
#### Note: These datasets should be used for research only.