Title: Machine Learning Approaches for Traffic Volume Forecasting: A Case Study of the Moroccan Highway Network
Authors:Abderrahim Khalifa, Younes Idsouguou, Loubna Benabbou, Mourad Zirari
(Submitted on 18 Nov 2017)
Abstract: In this paper, we aim to illustrate different approaches we followed while developing a forecasting tool for highway traffic in Morocco. Two main approaches were adopted: Statistical Analysis as a step of data exploration and data wrangling. Therefore, a beta model is carried out for a better understanding of traffic behavior. Next, we moved to Machine Learning where we worked with a bunch of algorithms such as Random Forest, Artificial Neural Networks, Extra Trees, etc. yet, we were convinced that this field of study is still considered under state of the art models, so, we were also covering an application of Long Short-Term Memory Neural Networks.
Submission historyFrom: Loubna Benabbou [view email]
[v1] Sat, 18 Nov 2017 00:26:44 GMT (33kb)
Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)
What is Traffic Volume Count ?
Traffic Data Collection is basic requirements for transport planning. Traffic Data forms an integral part of national economics and such knowledge is essential in drawing up a rational transport policy for movement of passengers and goods by both government and the private sectors. Traffic Volume Count is counting of number of vehicles passing through a road over a period of time. It is usually expressed in terms of Passenger Car Unit (PCU) and measured to calculate Level of Service of the road and related attributes like congestion, carrying capacity, V/C Ratio, identification of peak hour or extended peak hour etc. Traffic volume count or TVC is usually done as a part of transportation surveys, TVC can be classified or unclassified.
Need of Traffic Volume Count Survey
Traffic Volume Survey is an essential part of Town Planning, especially for a town planner. It includes counting the number of vehicles passing through a survey station. The study of Classified Traffic Volume Count is to understand factors that form the basis of:
a) Checking the efficiency/saturation of the road network by comparing current traffic volume with the calculated capacity or by identifying level of service
b) Establishing the use of the road network by vehicles of different categories, traffic distribution, PCU/vehicle value
c) Need of median shifting or road widening
Purpose of Traffic Volume Count
The purpose classified traffic volume count is to draw inferences on the basis of data collected. To provide possible solutions and improvement suggestion for the problem identified. The objectives covered in it includes identifying the hourly distribution of vehicles and peak hour, identify level of service and compare modal composition on different hierarchy of roads.
Methods of doing Traffic Volume Count
Traffic Volume Count can be done by various methods depending upon various factors like manpower available, budget, technology/instrument available, magnitude of traffic data required or to be collected which will then determine quality and type of vehicle classification to be adopted. Traffic counting falls in two main categories, namely: manual count and automatic count.Traffic data collection forms the integral part of traffic volume study as it provides the raw data and includes primary survey. The various types and methods used to collect traffic data not only provide a good and valuable coverage of the required traffic information. Different methods of traffic volume count are as mentioned below –
Duration and Interval of Traffic Counts
In order to predict traffic flow volumes that can be expected on the road network during specific periods, knowledge of the fact is required that traffic volumes changes considerably at each point in time. There are three important cyclical variations:
- Hourly pattern: the way traffic flow characteristic varies throughout the day and night;
- Daily Pattern: The day-to-day variation throughout the week
- Monthly and yearly Pattern: The season-to-season variation throughout the year.
When analysing the traffic one must also be aware of the directional distribution of traffic and the manner in which its composition varies as it is important to deal with tidal flow.
- Hourly patterns – Typical hourly patterns of traffic flow, particularly in urban areas, generally show a number of distinguishable peaks. Peak in the morning followed by a lean flow until another peak in the middle of the afternoon, after which there may be a new peak in the late evening. The peak in the morning is often more sharp by reaching the peak over a short duration and immediately dropping to its lowest point. The afternoon peak on the other hand is characterised by a generally wider peak. The peak is reached and dispersed over a longer period than the morning peak.
- Daily patterns – The traffic volume generally varies throughout the week. The traffic during the working days (Monday to Friday) may not vary substantially, but the traffic volume during the weekend is likely to differ from the working days on different type of roads and in different directions
- Manual Count
The most common method of collecting traffic volume data is the manual method of traffic volume count, which involves a group of people recording number of vehicles passing, on a pre-determined location, using tally marks in inventories. Raw data from those inventories is then organized for compilation and analysis. This method of data collection can be expensive in terms of manpower, but it is nonetheless necessary in most cases where vehicles are to be classified with a number of movements recorded separately, such as at intersections also in case where automatic methods cannot be used due to lack of infrastructure, necessary authorization etc.
2. Automatic Count
This method is employed in cases where manual count method is not feasible. Various instruments are available for automatic count, which have their own merits and demerits. Some of the widely used instruments are pneumatic tubes, inductive loops, weigh-in-motion Sensor, micro-millimeter wave Radar detectors and video camera. Both types of count can be classified or unclassified. Classified traffic volume count gives a better understanding of the types of vehicles which uses the road and can be used for number of other purposes apart from the transportation surveys. It can also be used for calculating the modal split of vehicles on the road. Unclassified traffic volume count is done where sufficient manpower is not available or the budget for the survey is low. This type of volume count does not give a good information about the road.
Some of the widely used instruments are –
i) Pneumatic tubes – These are tubes placed on the top of road surfaces at locations where traffic counting is required. As vehicles pass over the tube, the resulting compression sends a burst of air to an air switch.
ii) Inductive loops – Inductive loop detector consists of embedded turned wire. It includes an oscillator, and a cable, which allows signals to pass from the loop to the traffic counting device. Inductive loops are cheap, almost maintenance-free and are currently the most widely used equipment for vehicle counting and detection
iii) Weigh-in-Motion Sensor types – A variety of traffic sensors and loops are used to count, weigh and classify vehicles while in motion, and these are collectively known as Weigh In Motion (WIM) sensor systems. Some notable traffic sensors are:
- Bending Plates which contains strain gauges that weigh the axles of passing vehicles
- Capacitive Strip is a thin and long extruded metal used to detect passing axles. Capacitive strips can be used for both statistical data and axle configuration.
- Capacitive Mat functions in a similar manner as the capacitive strip but it is designed to be mobile and used on a temporary basis only.
- Piezo-electric Cable is a sensing strip of a metallic cable that responds to vertical loading from vehicle wheels passing over it by producing a corresponding voltage. The cable is very good for speed measurement and axle-space registration, and is relatively cheap and maintenance
iv) Micro-millimeter wave Radar detectors – Radar detectors actively emits radioactive signals at frequencies ranging from the ultra-high frequencies (UHF) of 100 MHz, to 100 GHz, and can register vehicular presence and speed and can be used determine vehicular volumes and classifications in both traffic directions..
v) Video Camera – Video image processing system utilize machine vision technology to detect vehicles and capture details about individual vehicles when necessary. The system is useful for traffic counting and give a +/- 3% tolerance, and is not appropriate for vehicular speed and their classification.
Factors to be considered while doing a traffic volume survey on mid block –
1. Surveyor should not affect the flow of traffic.
2. Survey station should be located at position where queuing do not take place.
3. Vehicles should be classified if possible as it saves time for Classified Traffic Volume Survey. Also classified results have many other application.
4. Safety of surveyor should be kept in mind and safe location should be selected. This becomes more important in rural area where carriageway is not well-defined.
5. Equipments used while automatic count should be placed such that they do not draw attention of driver.
Traffic Volume Survey can be done manually or by use of automatic methods depending upon various factors like manpower available, budget, technology/instrument available, magnitude of traffic data required
Read about: Trip Generation, Traffic Island & its Use