A new approach for modelling mixed traffic flow with motorized vehicles and non-motorized vehicles based on cellular automaton model

In this study, we provide a novel approach for modelling the mixed traffic flow. The basic idea is to integrate models for nonmotorized vehicles (nm-vehicles) with models for motorized vehicles (m-vehicles). Based on the idea, a model for mix traffic flow is realized in in the following two steps. At a first step, the models that can be integrated should be chosen. The famous NaSch cellular automata (NCA) model for m-vehicles and the Burgur cellular automata (BCA) model for nm-vehicles are used in this paper, since the two models are similar and comparable. At a second step, we should study coupling rules between m-vehicles and nm-vehicles to represent their interaction. Special lane changing rules are designed for the coupling process. The proposed model is named as the combined cellular automata (CCA) model. The model is applied to a typical mixed traffic scenario, where a bus stop without special stop bay is set on nonmotorized lanes. The simulation results show that the model can describe both the interaction between the flow of nm-vehicles and m-vehicles and their characters.

💡 Research Summary

The paper introduces a novel cellular automaton (CA) framework for simulating mixed traffic composed of motorized vehicles (m‑vehicles) such as cars and buses, and non‑motorized vehicles (nm‑vehicles) such as bicycles and pedestrians. The authors start from the observation that most existing traffic models treat these two classes separately, which limits their ability to capture the complex interactions that occur in real urban streets where both types coexist. To bridge this gap, they select two well‑established CA models that share a common discrete‑time, discrete‑space structure: the Nagel‑Schreckenberg (NaSch) model, here referred to as NCA, for motorized traffic, and the Burgur cellular automaton (BCA) for non‑motorized traffic. Because both models update all cells synchronously and use a single‑lane lattice, they can be integrated without major reformulation.

The integration proceeds in two stages. First, the two sub‑models are run in parallel on the same lattice, each governing its own set of agents. Second, a set of coupling rules is devised to govern the interaction between the two streams. The core of these rules is a lane‑changing mechanism that respects the physical priority that motorized vehicles typically have, while still allowing non‑motorized agents to avoid collisions. Specifically, an m‑vehicle may overtake or shift into a lane occupied by an nm‑vehicle only when the latter has moved at least two cells ahead, thereby ensuring a safe gap. Conversely, an nm‑vehicle encountering an approaching m‑vehicle will either stop or move laterally to the right, mimicking the evasive behavior of pedestrians and cyclists in real traffic.

To demonstrate the model’s capabilities, the authors construct a scenario that is common in many Asian cities: a bus stop without a dedicated bay is placed on a lane that is otherwise reserved for non‑motorized traffic. In the simulation, the bus is treated as an m‑vehicle that must stop for a prescribed dwell time, temporarily occupying cells on the non‑motorized lane. The coupling rules are activated whenever the bus approaches or departs from the stop, forcing nearby nm‑vehicles to either wait or detour. The lattice consists of 2,000 cells, with periodic boundary conditions, and the simulation runs for 10,000 time steps. Performance metrics such as average speed, flow (vehicles per time step), density, and the length of congestion clusters are recorded.

Results show that the presence of a stopping bus significantly reduces the flow of nm‑vehicles on the affected lane, while also inducing a spill‑over effect that slows down m‑vehicles on the adjacent lane as they either wait behind the bus or execute lane changes. The magnitude of the impact scales with the bus dwell time: longer stops lead to longer congestion clusters and a sharper drop in overall system throughput. A comparative experiment where a dedicated bus bay is provided demonstrates that the coupling rules become less active, and the mixed traffic operates much more efficiently, confirming the model’s sensitivity to infrastructure design.

The authors discuss the implications of these findings for traffic management. By capturing the micro‑level interactions between motorized and non‑motorized agents, the Combined Cellular Automata (CCA) model offers a quantitative tool for evaluating policies such as dedicated lanes, bus stop placement, and signal timing. The paper also acknowledges limitations: the current implementation assumes a single‑dimensional, static lane configuration and a limited set of vehicle types (cars, buses, bicycles, pedestrians). Extending the framework to multi‑lane networks, complex intersections, dynamic signal control, and emerging micro‑mobility modes (e‑scooters, shared bikes) would increase its realism. Moreover, calibration against empirical traffic data is necessary to validate parameter choices and to assess predictive accuracy.

In conclusion, the study demonstrates that integrating NCA and BCA through carefully designed coupling rules yields a flexible and realistic mixed‑traffic simulation platform. The CCA model successfully reproduces characteristic phenomena such as temporary flow reductions, queue formation, and lane‑changing behavior observed in real urban environments. The authors propose future work that includes field data validation, incorporation of stochastic driver behavior, and application of the model to urban planning scenarios, thereby positioning the CCA as a valuable decision‑support tool for modern, multimodal transportation systems.