Extensive research on point-detector-based automatic traffic-impeding incident detection indicates the potential superiority of neural networks over conventional approaches. All approaches, however, including neural networks, produce detection algorithms that are location specific—that is, neither transferable nor adaptive. A recently designed and ready-to-implement freeway incident detection algorithm based on genetically optimized probabilistic neural networks (PNN) is presented. The combined use of genetic algorithms and neural networks produces GAID, a genetic adaptive incident detection logic that uses flow and occupancy values from the upstream and downstream loop detector stations to automatically detect an incident between the said stations. As input, GAID uses modified input feature space based on the difference of the present volume and occupancy condition from the average condition for time and location. On the output side, it uses a Bayesian update process and converts isolated binary outputs into a continuous probabilistic measure—that is, updated every time step. GAID implements genetically optimized separate smoothing parameters for its input variables, which in turn increase the overall generalization accuracy of the detector algorithm. The detector was subjected to off-line tests with real incident data from a number of freeways in California. Results and further comparison with the McMaster algorithm indicate that GAID with a PNN core has a better detection rate and a lower false alarm rate than the PNN alone and the well-established McMaster algorithm. Results also indicate that the algorithm is the least location specific, and the automated genetic optimization process makes it adapt to new site conditions.
[1]
Fred L. Hall,et al.
DISTINGUISHING BETWEEN INCIDENT CONGESTION AND RECURRENT CONGESTION: A PROPOSED LOGIC
,
1989
.
[2]
Bhagwant Persaud,et al.
CONGESTION IDENTIFICATION ASPECTS OF THE MCMASTER INCIDENT DETECTION ALGORITHM
,
1990
.
[3]
Baher Abdulhai,et al.
A neuro-genetic-based universally transferable freeway incident detection framework
,
1996
.
[4]
T. Cacoullos.
Estimation of a multivariate density
,
1966
.
[5]
D. F. Specht,et al.
Enhancements to probabilistic neural networks
,
1992,
[Proceedings 1992] IJCNN International Joint Conference on Neural Networks.
[6]
Yong Shi,et al.
ON-LINE TESTING OF THE MCMASTER INCIDENT DETECTION ALGORITHM UNDER RECURRENT CONGESTION
,
1993
.
[7]
Timothy Masters,et al.
Advanced algorithms for neural networks: a C++ sourcebook
,
1995
.
[8]
D. F. Specht,et al.
Experience with adaptive probabilistic neural networks and adaptive general regression neural networks
,
1994,
Proceedings of 1994 IEEE International Conference on Neural Networks (ICNN'94).
[9]
E. Parzen.
On Estimation of a Probability Density Function and Mode
,
1962
.
[10]
Baher Abdulhai,et al.
PERFORMANCE OF ARTIFICIAL NEURAL NETWORKS FOR INCIDENT DETECTION IN ITS
,
1995
.
[11]
Baher Abdulhai,et al.
Preprocessor Feature Extractor and Postprocessor Probabilistic Output Interpreter for Improved Freeway Incident Detection
,
1999
.
[12]
Timothy Masters,et al.
Advanced algorithms for neural networks: a C++ sourcebook
,
1995
.