Close following behavior: Testing visual angle car following models using various sets of data

Visual angle car following models have previously been proposed in modeling drivers’ “close following” behavior particularly in traffic simulation models. These models use relative speeds and relative spacings between successive vehicles, but they are unique in including the width of vehicles when determining the safe distance between the leading vehicle and its follower. This paper focuses on examining the validity of the visual angle model assumptions; especially the ability of such models in representing the effect of heavy goods vehicles, when such models are used in traffic simulation applications. Two types of data have been used for this purpose. Published real traffic data from instrumented vehicles (both from the USA and Germany) have been used to check the capability of these models to replicate real traffic movements but only when both leading and following vehicles are considered to be “small cars”. Another type of data has been used based on two motorway sections in the UK, namely the M25 (London outer ring road) and a section of the M42 “Managed Motorway” (Birmingham outer ring road). Over 4 million cases of individual vehicles have been abstracted from inductive loop detectors installed on these motorways. The sets of data were then filtered and analyzed to examine the following distance according to the type of leading vehicles (i.e. “cars” or heavy goods vehicles – “HGVs”). The results showed that while visual angle car following models may successfully replicate real traffic movements based on the instrumented vehicles data from Germany and the USA, the assumption of leaving larger following distance if the leader is an HGV is found to be not the case for the majority of UK drivers. This will have a negative impact on the ability to use visual angle car following simulation models to represent UK real traffic behavior.

Research highlights
► This paper examined the ability of visual angle car following model using instrumented data.
► A substantially larger database from the UK was used to test vehicles’ close spacings.
► The effect of vehicles’ type on close following distance was studied.
► No evidence was found that vehicles’ type has any significant effect on close following distance.
► Therefore, the validity of visual angle models to represent real traffic behavior is in question.