VASIC, RUSKIN: Discrete simulation model for traffic including bicycles on urban networks THROUGHPUT AND DELAY IN A DISCRETE SIMULATION MODEL FOR TRAFFIC INCLUDING BICYCLES ON URBAN NETWORKS

The 'greening' of transport is an ongoing concern in today's cities, with many struggling to reach carbon emission targets. One relevant area for effecting change is encouragement and facilitation of alternative, non-motorised transport modes, such as cycling and walking. Understanding the dynamics of non-motorised flows and their interaction with motorised traffic in an urban context is fundamental to exploring these alternatives and has been a recent focus of much technical (and social) research. In this paper, we describe the development of an agent-based simulation framework for mixed traffic on urban networks and its application to a basic network scenario, in which heterogeneity takes the form of lane-sharing by bicycles and cars. Performance is analysed in terms of throughput for different network and signalisation parameter values. Introduction The study of the physical properties of traffic is an integral part of transportation science, contributing information that forms the basis for transportation related management and policy decisions. Accordingly, he growing relevance of non-motorised modes of transport, which have beneficial characteristics with respect to the environment, health and society [1] [2], has created a need for the investigation of heterogeneous flow dynamics with focus on alternative modalities, such as walking and cycling. Of particular interest are the properties of motorised/non-motorised mixed urban traffic, as non-motorised modes are predominantly used in urban settings. While non-motorised traffic and the motorised-non-motorised mix have, as topics, had their place in traffic flow science in the past, in recent years, with the mounting of environmental concerns in the world and, in parallel, with the fast motorisation of developing countries, they have gained increasing attention of the traffic science community. Because of the many forms that non-motorised traffic integration with the overall road transportation system can take, the study of such traffic has been undertaken for a variety of specific conditions. For the purpose of dedicated bicycle facility design, there have been several studies, reviewed in [3], on track and lane capacity and level of service. Two recent bicycle-only flow models were developed using cellular automata [3][4]. Models for interactions between bicycles and motorised traffic in adjoining lanes have been modelled in [5][6] and [7], the focus in the former two being on conflicts at intersections and that of the latter on the inadvertent invasion of neighbouring, other modality lanes. The flow of fully mixed traffic, which occurs in the absence of lane discipline, was modelled in [8] and [9]. Traffic flow on networks, which is additionally impacted by topology, has also been modelled for the non-motorised and mixed cases. Reviews of a number of continuous-space simulation models, predominantly for traffic including bicycles without lane discipline, are given in [10], while [11] presents another complex model that includes bicycle flows and their various interactions on a network. In addition, some commercial offerings, such as VISSIM [12], simulate bicycle behaviour as part of their functionality. In this paper, we describe a cellular automaton based network traffic simulation model and VASIC, RUSKIN: Discrete simulation model for traffic including bicycles on urban networks 31st August – 1st September, University College Cork Proceeding s of the ITRN2011 corresponding implementation framework that we have developed for the study of bicycle traffic within the wider context of heterogeneous vehicle flows in an urban setting. Our focus is on conditions similar to those prevailing in Dublin, where lanes are shared between motorised traffic and bicycles, lane sharing is based on 'positional discipline' (i.e. bicycles keep to the left and motorised vehicles to the right of the lane) and bicycle-dedicated controls are scarce. We apply the model to a small regular network and present the results of performed simulations in terms of overall flows and average trip times for bicycles and cars separately. This example, while simple, demonstrates the ultimate purpose of our model, which is to analyse general road traffic network properties, such as topology, geometry, rules of behaviour and signalisation, with respect to bicycles. The following section, “Model”, briefly introduces the model and the implementation framework. Section “Network simulation” describes the simulated network and the results of a number of different simulation scenarios. The final section, “Conclusion”, summarises the conclusions drawn from the simulation results and introduces the questions that will be tackled by continuation of this research. Model Since first used by Nagel and Schreckenberg [13] for the simulation of a one-way flow of cars, cellular automata have been widely applied for the reproduction of traffic behaviour. Simulations based on cellular automata are fairly simple to implement and offer the advantage of computational efficiency. A cellular automata model is time and space-discrete. This means that time is represented by discrete moments called time steps, which occur at regular intervals, while space is represented by a grid of cells, which can either be occupied or unoccupied at any time step. Vehicles were initially represented as occupying one and one only cell, however, for reasons of modelling vehicles of different sizes and shapes (e.g. [9]), and model refinement (e.g. [14]), multiple cell occupancy by individual vehicles was introduced into models. Inversely, in [4] multiple bicycles can occupy a single cell. In our model, which was introduced in [15], we use separate but overlapping cellular systems to solve both the problem of differently sized vehicles and that of intersection conflicts in the model. Each type of vehicle moves on a lattice of cells that allow it to occupy one and exactly one whole cell at any one time, i.e. a lattice of cells that are of that vehicle type's size. Intersections are represented as a collection of routes that can be taken by a vehicle through that intersection and each one modelled as a one-dimensional cellular automata space. These routes cross each other as the traverse the intersection, which, in addition to the superimposed cellular spaces of different granularity, is the other cause of overlap between cells. The overlap between cells and the accompanying concept of impingement, which is the indicator of whether a cell is available for occupation by a moving vehicle, is one of the two main differences between our model and previous cellular automata models of traffic flow. While these previous models (cf. [13]) would inspect cells for whether they are occupied or not when determining how to move, in our model vehicles check cells for impingement. The other important difference between previous models and ours is the spatial aspect of conflict expression, which in our model follows naturally from the spatial representation of an intersection: a conflict is described in terms of which sections (i.e. which cells) of overlapping routes constitute it. This way of representing conflicts allows for the re-use of conflict resolution rules, since a conflict is always expressed in the form of two cell ranges: the cells of one route and the cells of the conflicting route that are within the 'conflict zone'. We use the Nagel-Schreckenberg [13] rules as a base for the behaviour of vehicles. These are: 1) Acceleration: if vi < vMAX and vi < di , vi vi + 1, 2) Slowing (due to cars ahead): if di < vi , vi di, 3) Randomisation: with probability pR , vi vi – 1, 4) Vehicle motion: each vehicle is advanced vi cells, Proceeding s of the ITRN2011 31st August – 1st September, University College Cork VASIC, RUSKIN: Discrete simulation model for traffic including bicycles on urban