تحلیل تطبیقی روش‌های تخصیص ترافیک قطعی و تصادفی در برآورد حجم تردد

از آنجا که نتایج مدل تخصیص ترافیک، خروجی اصلی مطالعات برنامه‌ریزی حمل‌ونقل و پایه‌ی اصلی تصمیم گیری برای توسعه‌های آتی محسوب می‌شود، دقت برآورد این مدل‌ها اهمیت ویژه‌ای دارد. ولی مقایسه‌ی این مدل‌ها و روش‌های حل آنها در برآورد احجام واقعی و مشخصات کلی عملکرد شبکه معابر کمتر مورد توجه قرار گرفته است. هدف اصلی مقاله‌ی حاضر تحلیل تطبیقی و کمی نتایج روش‌های مختلف تخصیص ترافیک در برآورد و بازسازی احجام مشاهده شده است. فرآیند تخصیص از ساده‌ترین روش‌ها شامل همه یا هیچ شروع شده، با ارائه‌ی اصول، قوانین و فرض‌های مکمل نظیر واردراپ توسعه یافته و با ورود مفاهیمی نظیر منطق فازی تکامل پیدا کرده است. در این پژوهش روش‌های تخصیص متنوعی نظیر همه یا هیچ، جزئی، تخصیص تصادفی، تعادل کاربر، تعادل تصادفی کاربر و بهینگی سیستم مورد بررسی قرار گرفته‌اند. نتایج اجرای این روش‌ها در مورد شبکه معابر شهر قزوین (به عنوان نمونه موردی) نشان می‌دهد که انواع روش‌های تعادلی در برآورد شاخص‌های کلان شبکه و همچنین دقت تخمین حجم تردد در کمان‌ها، علی رغم فرض‌های متنوع، تفاوت معنی‌داری ندارند (ضریب خوبی برازش مشاهده و برآورد حدود 0.88). ولی روش‌های ساده که رابطه‌ای بین حجم و زمان سفر در نظر نمی‌گیرند، به صورت محسوس نتایج متفاوتی نسبت به روش‌های تعادلی ارائه می‌کنند (ضریب خوبی برازش حدود 0.70). با توجه به توزیع احتمال خطای برآورد حجم که دارای میانگین و انحراف معیار قابل توجهی است (حدود 20 درصد)، باید در کاربرد نتایج روش‌های تخصیص دقت شود.

Comparative Analysis of Deterministic and Stochastic Traffic Assignment Methods for Estimating the Actual Traffic VolumeCase Study: Qazvin City

Results of traffic assignment models are the main output of transportation planning studies and decision making for future developments is based on these results. Therefore, accuracy of these models is very important. Despite the mentioned importance, comparing the models and their solving methods to estimate actual traffic volume and network performance measures is rarely considered in previous studies. The traffic assignment process has started from the simplest methods like AllorNothing, then it has developed using the rules and supplement assumptions such as Wardrop principles and finally it has evolved by concepts such as Fuzzy theory. Traffic assignment models can be categorized by various factors into several groups: deterministic vs. stochastic, congestion considering vs. unconstrained capacity and being equilibrium or not. The main goal of this paper is a comparative and quantitative analysis of various traffic assignment methods to estimate the observed traffic volumes. In this regard, the main questions that this study seeks to answer is as follows: 1 Do the results of various traffic assignment methods have a significant difference in terms of overall network indices? 2 Is there a significant difference in the accuracy of traffic volume estimation in various traffic assignment methods? In this study various traffic assignment methods such as AllorNothing, Incremental, Stochastic, User Equilibrium, Stochastic User Equilibrium and System Optimum have been examined. To compare the results of traffic assignment methods, in addition to estimated link volumes, various performance measures such as vehiclekilometers traveled, vehiclehours traveled, fuel consumption and air pollutants emission are also used. In this regard the city of Qazvin is selected as a case study. This city has more than 400 thousands inhabitant, near 46 square kilometers area, 113 traffic analysis zone (TAZ) and its network has 2300 directional links and 1200 nodes. The results of applying these methods in Qazvin city network show that various traffic assignment methods based on User Equilibrium, despite different assumptions, have no significant difference in estimating the overall network performance measures as well as estimating traffic volume in links (correlation between estimated and actual link volumes using all of these models is approximately 0.88). But the other methods, which do not consider equilibrium assumption and volumedelay functions, produce different results (correlation between estimated and actual link volumes using all of these models is approximately 0.70). Although estimated link volumes in some of traffic assignment models are significantly different, overall network performance measures are approximately the same. In all of assignment models the differences between estimated and actual link volumes in average are high which are not negligible (approximately 20 percent). In addition to high average error in estimating link volumes, the distribution of these errors has significantly high standard deviation (approximately 20 percent). In spite of different and complicated assumptions, models and solving algorithms in various traffic assignment methods, on basis of KolmogorovSmirnov (KS) test results, the distribution of links volume estimation error is not significantly different. According to this fact, it seems that should be careful in using the results of traffic assignment models to compare and assess minor network improvement alternatives, such as changing conventional streets function to pedestrian streets, upgrading intersections to interchanges, cross section widening, traffic signals optimization and changing traffic direction in streets.