Motorway Bottleneck Probability Estimation in Connected Vehicles Environment Using Speed Transition Matrices

Sensors (Basel). 2022 Apr 6;22(7):2807. doi: 10.3390/s22072807.


Increased development of the urban areas leads to intensive transport service demand, especially on urban motorways. To increase traffic flow and reduce congestion, motorway traffic bottlenecks caused by high traffic demand need to be efficiently resolved using Intelligent Transport Systems services. Communication technology development that supports Connected Vehicles (CVs), which act as an active mobile sensor for collecting traffic data, provides an opportunity to harness the large datasets to develop novel methods regarding traffic bottlenecks detection. This paper presents a speed transition matrix based model for bottleneck probability estimation on motorways. The method is based on the computation of the speed at the vehicle transition point between consecutive motorway segments, which forms a traffic pattern that is represented using transition matrices. The main feature extracted from the traffic patterns was the center of mass, whose position is used as an input to the fuzzy-based system for bottleneck probability estimation. The proposed method is evaluated on four different simulated motorway traffic scenarios: (i) traffic collision site, (ii) short recurring bottleneck, (iii) long recurring bottleneck, and (iv) moving bottleneck. The method achieves comparable bottleneck detection results on every scenario, with a total accuracy of 92% on the validation dataset. The results indicate possible implementation of the method in the motorway traffic environment with a high CVs penetration rate using them as the sensory input data for the control systems based on the machine learning algorithms.

Keywords: bottleneck detection; bottleneck probability; connected vehicles; fuzzy-based bottleneck probability; motorway bottleneck; speed transition matrix; traffic simulation.

MeSH terms

  • Accidents, Traffic* / prevention & control
  • Algorithms
  • Automobile Driving*
  • Data Collection
  • Probability