See Events

PhD Defence Erwin Bezembinder | Junction design rules: Improving junction design choices in urban traffic networks

Junction design rules: Improving junction design choices in urban traffic networks

Due to the COVID-19 crisis the PhD defence of Erwin Bezembinder will take place (partly) online.

The PhD defence can be followed by a live stream.

Erwin Bezembinder is PhD student in the research group Transport Engineering and Management (TEM). His supervisor is E.C. van Berkum and his co-supervisor is L.C.C. Wismans, both from the Faculty of Engineering Technology (ET).

Special note: Erwin Bezembinder and Marie Jose Olde Kalter both have their defence on November 5, one after the other (respectively 12.45 and 14.45). They both belong to the faculty Engineering Technology, department Civil Engineering. What makes this such a special occasion is the fact that they are a married couple.

Transport planners and engineers frequently face the challenge to determine the best design for a specific junction. Junction design encompasses the choice for the main type such as equal, priority, signalised or roundabout and choices regarding the layout such as the number and configuration of the lanes, the central reservation width, the number and type of lanes on a roundabout, the facilities for bicycles and pedestrians, and the signal control settings. Junction design is determinative for the effects of traffic in urban areas, both on local and network level. The choice for a particular junction design leads to certain levels of operation, safety and environmental impact on the junction and on other junctions and the road network. For road authorities it is important to find the junction design which minimises one or more negative effects, corresponding to the related policy objectives, both on local and network level. The contemporary junction design assessment process incorporates two main deficiencies. One regarding deficient identification rules used in the first stage of the assessment process and one regarding insight in and tools to control network effects of local selection methods used in the last stage of the process. This thesis provides solutions for these two deficiencies.

Alternative identification

This thesis provides junction design rules, as well as a suitable approach to determine and evaluate them, which can be used to identify junction design alternatives which benefit most to minimising the negative effects of traffic on isolated junctions. In order to determine the new identification rules, datasets were generated containing the operational, safety and environmental performances for a wide range of junction design alternatives and traffic flow patterns. Pareto optimal sets of junction design alternatives were determined for each traffic flow pattern in the data. Decision tree methods were used in order to derive new rules from the data. Existing and new junction design rules were evaluated for two cases. The main conclusions concerning alternative identification are:

  • Results showed that for about 90% of the situations the Pareto optimal set could be predicted by the new rules, whereas the existing rules hardly reached 35%.
  • The junction design alternatives identified by the new rules provide better performances (i.e. less negative effects). For the Dutch case, the operational performances are reduced with 28% when compared to the existing rules. Safety and environmental performances are reduced with respectively 26% and 60%.
  • The new rules provide smaller sets of viable junction design alternatives. Existing rules are less discriminating, resulting in large set sizes. A 41%-52% reduction of set size can be achieved by the new rules, which might imply a similar reduction in cost in later stages of the junction design assessment process.
  • It is possible to generate accurate and comprehensible junction design rules that can be included in junction design manuals. The determined rules only require four volume based independent variables as input to provide a (Pareto optimal) set of viable junction design alternatives.
  • The suitability of the modelling approach used is demonstrated. The approach (including a new multi-label decision tree method) can easily be used to generate new or update existing junction design rules.

Network effects

This thesis also provides insights in the network effects of local selection methods and provides specific rule based selection methods which can be used to determine network strategies for junction design in order to minimise the negative effects of traffic on a network level. In order to obtain insight in the network effects of junction selection methods, the operational, safety and environmental performances of different selection methods were determined and analysed using various modelling approaches incorporating route choice behaviour. Models for traffic assignment and local and global optimisation (formulated as a specific case of the Road Network Design Problem) were used to evaluate existing and new (rule based) selection methods for two cases. The main conclusions concerning network effects are:

  • Results showed that a substantial reduction of negative effects of traffic can be achieved by optimising the junction designs. For the Delft case, a 37% reduction of total loss time, a 6% reduction of the total number of fatal-and-injury crashes per year and a 1% reduction of NOx emissions could be obtained in comparison with the current situation.
  • The network effects of local selection criteria using delay, especially those using volume weighted average delay, are very close (1%-point) to those of the global (operational) optimum situation, for operation, safety and emissions.
  • The solutions from the operational optimum and the operationally best performing local criteria contain relatively few signalised junctions and far more priority junctions.
  • The local selection criterion based on sustainable safety on the other hand, performs poorly. First and foremost for operation, but surprisingly, due to the route choice effects, the results for safety are also below expectations.    
  • Applying the local selection criteria on a limited number of junctions each time (e.g. year), generally leads to poorer network performances, also when eventually the same number of junctions has been evaluated. 
  • Based on data retrieved from the global optimisation, rule-based selection criteria were determined by using decision tree methods. The resulting (non-performance requiring) volume based rules can be used to evaluate local selection criteria for existing urban networks without the necessity to define and evaluate junction design alternatives for all junction locations in the network.     

Final remark

An important conclusion from this research is that a substantial reduction of negative effects of traffic can be achieved by changing the junction designs. Nowadays, a lot of effort and money is invested in network optimisation using traffic management measures such as smart or intelligent traffic lights, which are flexible and effective tools in order to manage traffic flows. However, one should keep in mind that the criteria for selecting a specific junction design alternative have similar or even bigger impacts on traffic flows, safety and air quality in urban networks.