Information Technology Reference
In-Depth Information
5
Conclusions
This paper discussed the development of the FLM for estimating the 85
th
percentile
speed based on six road's attribute data. The advantage of fuzzy logic is its ability to
address the uncertain nature of human thinking (perception). The same road (road
attribute data) can be perceived differently by different drivers and choose their oper-
ating speed accordingly. The other advantage is the using the neuro-fuzzy which can
be utilized to automate the development of the knowledge base.
The FLMs are widely known for describing the vagueness and nonlinearity in the
human behavior relationships between inputs and output. However, such models are
generally only valid in situations for which data are available to calibrate the model. If
the FLM is to be used to assess the choice behavior that is not covered in the data for
calibration, the applicability of the model for estimating the 85
th
percentile speed
might be questionable. As such, the data for calibration should thoroughly cover the
entire range of (input and output) variables for better and more accurate estimation.
Identifying and setting appropriate posted speed limit for a given road segment is a
complex task which involves studying the speed behavior pattern of the drivers, the
characteristics of road environment, road geometry, etc. This study focused on only
one aspect; the drivers' speeding behavior based on the basic road characteristics, the
traffic intensity and pedestrian activities for a very limited number of road segments.
One may argue the necessity to develop such models while such 85
th
percentile
speed can be actually measured in the field. In response to such argument is that
tremendous savings in the resources (that would be needed to carry on actual field
survey measures over an entire network) can be materialized. It is envisioned that
these models can be developed with a reasonable representative sample of road seg-
ments in a typical network. The derived models can then be validated and subsequent-
ly applied to the entire network.
Keeping in mind the limited data set used in the study (due to the resources con-
straints), that likely contributes to deficiencies in representing the various road cha-
racteristics and environmental factors (with only few data points); it is legitimate to
assume that the richness in data collection will ultimately lead to better more statisti-
cally significant models. Along this line, it is suggested that a systematic sampling
approach should be adopted in selecting the road segments to include in the data set to
use for models' calibration. The principles of the minimum sample size should be
observed. It is suggested that a stratified sampling procedure to be used in selecting
the road segments for spot speed field observations. All the network roadway seg-
ments may be stratified based on their intrinsic characteristics of posted speed, length,
traffic volume, pedestrian intensity, etc. A representative stratified sampling proce-
dure with a minimum sample size according to a pre-specified confidence level and
interval should be observed in generalizing the fuzzy logic modeling approach.
Acknowledgements.
This research is part of M.Sc. thesis entitled 'Assessing the
Methodology of Setting Posted Speed Limit in Al Ain-UAE' funded by Roadway,
Transportation and Traffic Safety Research Center, United Arab Emirates University.
