Estimating traffic capacity for weaving segments using neural networks technique
Introduction
The capacity of a weaving segment is defined according to the Highway Capacity Manual (HCM) 2000 [16] as “any combination of flows that causes the density to reach LOS E/F boundary condition of 27.0 pc/km per lane for freeways or 25.0 pc/km per lane for multilane highways.” Many variables contribute to capacity estimates at weaving segments, such as segment configuration, number of lanes, free flow speed on the freeway or multilane highway, weaving length, and volume ratio. However, since it was not possible to develop a simple closed-form solution for capacity, HCM 2000 suggests using an iterative process to estimate capacity on weaving segments.
Exhibit 24-8 of the HCM 2000 tabulates capacities for a number of situations under base conditions for freeway facilities. Straight-line interpolation is required, as a rough estimate, for any intermediate value or values of the segment conditions. Therefore, for any given set of conditions, the algorithms described in exhibit 24-8 must be solved iteratively to find capacity using a properly programmed spreadsheet. For prevailing conditions, capacities in exhibit 24-8 are then adjusted for heavy-vehicle presence and driver familiarity.
Section snippets
Literature review
Most previous related research focused on predicting speed and level of service (LOS) for weaving segments [8], [10], [12], [13], [14]. The estimation of capacity along weaving segments has not been well investigated, Cassidy and May [2] proposed an analytical technique for estimating capacity, while Lertworawnich and Elefteriadou [9] developed a method for estimating capacities of type B weaving areas based on gap acceptance and linear optimization. Most past research focused on improving the
Motivations and objectives
As indicated above, most studies in the past focused on the development of regression-type models trying to capture combined effects of weaving flows without explicitly addressing the interaction of the contributing factors, or account for the associated uncertainty.
The HCM 2000 procedure for estimating the capacity of weaving segments requires iterative process using a properly programmed spreadsheet to interpolate the tabulated capacities in exhibit 24-8 to account for unlisted base
Weaving data and models formulation
The data in exhibit 24-8 of the HCM 2000 is used in this study. The data includes capacity of weaving segments referred herein as the criterion variable (Ci) for different combinations of five explanatory variables (i.e. segment configuration, number of lanes, free flow speed on the freeway or multilane highway, weaving length, and volume ratio). The three major types of weaving configurations are considered. Types A, B, and C are defined based on the number of lane changes needed for each of
Neural networks model
Artificial neural network techniques are computational tools with different structures; attempt to simulate the architecture of human brain and neurons system. Fig. 1 shows a typical neural network with three layers, namely input, hidden, and output layer, each layer contains a specified number of nodes based on the number of variables in the input and output layers. The nodes across layers are interconnected, and random initial weights are assigned to connections and output is calculated.
Results and models validation
Two techniques are used to predict capacity of weaving segments using the data in exhibit 24-8 (HCM 2000), namely linear regression and neural networks. The reduced models as shown in Table 3 are developed using LR technique only, while different full models are developed using both techniques (LR and NNT). The results of the full models are shown in Fig. 3, Fig. 4, Fig. 5, Fig. 6, Fig. 7, and discussed herein. Fig. 3, Fig. 4, Fig. 5 show the prediction ability of the models in Table 4 for
Conclusions
A comparative study is carried out for the use of multi-layer feed-forward neural networks versus linear regression in predicting capacity of weaving segments. The current practice (HCM 2000) estimates capacity from exhibit 24-8 either directly for listed values of the explanatory variables, or by interpolation for unlisted values of the explanatory variables. The interpolation is done by iterative process using a properly programmed spreadsheet. This study proposes two alternative methods to
References (18)
- W.H. Awad, B.N. Janson, Prediction models for truck accidents at freeway ramps in Washington state using regression and...
- M. Cassidy, A. May, Proposed analytical technique for estimating capacity and level of service of major freeway weaving...
- J. Fazio, N.M. Rouphail freeway weaving sections: comparison and refinement of design and operations analysis...
- J. Fazio, N.M. Rouphail, Conflict simulation in INTRAS: application to weaving area capacity analysis, Transportation...
- K. Fitzpartric, L. Nowlin, One-sided weaving operations on one-way frontage roads, Transportation Research Board Annual...
- V.E. Frederickson, M.A. Ogden, Proposed analytical technique for analyzing Type A weaving sections on frontage roads,...
- P.H. Goodman, F.E. Harrell Jr., Neural networks: advantages and limitations for biostatistical modeling,...
- Jack E. Leisch & Associates, Completion of procedures for analysis and design of traffic weaving sections, Final...
- P. Lertworawnich, L. Elefteriadou, Capacity estimation for type b weaving areas based on gap acceptance, Transportation...
Cited by (12)
Research on the weaving area capacity of freeways under man–machine mixed traffic flow
2023, Physica A: Statistical Mechanics and its ApplicationsApplication of neural networks and fuzzy logic models to long-shore sediment transport
2011, Applied Soft Computing JournalCitation Excerpt :Then, the methods have been extended to many other disciplines in civil engineering such as structural engineering, transportation, water network distributions, etc. Estimating traffic capacity [13], predicting frequency of annular elliptic and circular plates [14], optimizing structural topology [15], determining the characteristics of the multi-storey structure and of the earthquake motion [16] and modeling shear strength of SFRC corbels without stirrups [17] are among the published works. In general, soft computing methods provide a better alternative, either substitutive or complementary, to traditional computational schemes of statistical regression, time series analysis, pattern matching, and numerical methods in these areas.
Application of intelligent systems for specification of automotive equipments using LEDs
2005, Applied Soft Computing JournalModelling Acceleration Lane Length in CAV Lane Merge Area Based on Breakdown Probability
2022, Proceedings of SPIE - The International Society for Optical EngineeringAn intelligent location and state reorganization of traffic signal
2020, Geodesy and Cartography (Vilnius)Efficient goods inspection demand at ports: a comparative forecasting approach
2019, International Transactions in Operational Research