Micro-simulation of Traffic
Learn the foundational theory behind microscopic traffic simulation, its mathematical drivers (car-following and lane-changing models), and how engineers calibrate and validate these tools for real-world application.
Microscopic Traffic Simulation
The use of computational software to model the movements and interactions of individual vehicles (or pedestrians) within a transportation network over time, based on behavioral algorithms. Unlike macroscopic models that deal with aggregate averages (flow, density), microsimulation tracks the specific position, speed, and acceleration of every single entity every fraction of a second.
Core Behavioral Models
The "brain" of every simulated vehicle.
1. Car-Following Models
The most critical algorithm in microsimulation. It dictates how a simulated driver (the follower) reacts to the speed and position of the vehicle directly in front of them (the leader) within the same lane.
- The Stimulus-Response Framework: The basic premise is that a driver's acceleration or braking (response) is proportional to the relative speed between them and the leader (stimulus), and inversely proportional to their distance apart.
- The Wiedemann Model (Psycho-physical): Used prominently in VISSIM. It assumes drivers cannot perfectly perceive small changes in speed or distance. They only react when these differences cross a specific perceptual threshold. The model defines four driving regimes:
- Free driving: Unconstrained by the leader.
- Approaching: Reacting to a slower leader by decelerating.
- Following: Maintaining a safe distance with minor speed oscillations.
- Braking: Emergency deceleration if the distance becomes dangerously small.
2. Lane-Changing Models
This algorithm determines when and why a simulated vehicle decides to move from one lane to another.
- Discretionary Lane Changes: The driver wants to change lanes to improve their speed or pass a slower vehicle.
- Mandatory Lane Changes: The driver must change lanes because their current lane is ending, or they need to reach a specific turning lane for their route.
- Gap Acceptance Theory: Before executing any lane change, the model checks if the available gap in the target lane is larger than the driver's accepted "Critical Gap." It must ensure the change won't cause the approaching vehicle in the target lane to brake dangerously hard.
The Simulation Process
The structured workflow for developing a reliable microsimulation model.
Procedure
1. Network Building (Coding)
The engineer constructs the physical environment within the software (e.g., VISSIM, AIMSUN, Paramics, or SUMO). This involves drawing the road links, defining lane widths, setting speed limits, and inputting precise traffic signal timing plans.
2. Demand Input
The engineer inputs the traffic volume. In microsimulation, this is often done using a dynamic Origin-Destination (O-D) matrix, which tells the software how many vehicles are traveling from Zone A to Zone B over a specific time profile (e.g., 15-minute intervals), rather than just static hourly link counts.
3. Error Checking
A visual review of the animation to ensure there are no glaring coding errors (e.g., cars driving through buildings, signals stuck on red, or impossible lane configurations causing immediate gridlock).
4. Calibration
The single most critical and difficult step. The software's default behavioral parameters (like average standstill distance, reaction time, or aggressive lane-changing behavior) must be iteratively adjusted so that the model accurately replicates local real-world driving behavior. If calibration is skipped, the model's outputs are meaningless.
5. Validation
The calibrated model is tested against an independent set of real-world field data (e.g., travel times, queue lengths, or bottleneck locations) that was not used during the calibration phase. If the model's outputs closely match the validation data, it is considered robust and ready for forecasting.
6. Alternative Analysis
The validated model is used to test various "what-if" scenarios (e.g., adding a new interchange, implementing a transit-only lane, or altering signal phasing) to predict future performance before millions of dollars are spent on construction.
Advantages and Limitations
When to use microsimulation versus traditional Highway Capacity Manual (HCM) methods.
Advantages over Macroscopic Models
- Complex Geometries: Excellent for analyzing closely spaced intersections, complex interchanges (like diverging diamond interchanges), and roundabouts where traditional equations fail.
- Queue Spillback: Accurately models the cascading effects of a queue from one intersection backing up and blocking an upstream intersection (which HCM struggles with).
- Multimodal Interaction: Can visually and mathematically simulate the interaction between cars, heavy trucks, buses, light rail, and pedestrians in a single unified environment.
- Visual Communication: Produces high-quality 3D animations that are incredibly powerful for communicating complex engineering concepts to the public and non-technical politicians.
Limitations
- Extremely Data and Time Intensive: Building, calibrating, and validating a model requires massive amounts of field data and hundreds of hours of engineering time.
- Stochastic Nature: Because arrivals and driver behaviors are randomized, a single simulation run is not statistically valid. The engineer must run the simulation multiple times (often 10 to 20 times) using different random "seed" numbers and average the results to account for daily variation.
- The "Black Box" Effect: If the underlying algorithms and parameters are not well-understood by the user, it is easy to inadvertently manipulate the model to produce a desired (but inaccurate) result.
Key Takeaways
- Microsimulation models the specific, fraction-of-a-second behavior of individual vehicles using car-following and lane-changing algorithms.
- The Wiedemann Model is a prominent psycho-physical car-following model based on human perceptual thresholds.
- Calibration (adjusting behavioral parameters to match local driving habits) is the most critical step; uncalibrated models produce invalid results.
- Due to its stochastic (randomized) nature, a microsimulation model must be run multiple times with different random seeds to generate statistically reliable outputs.
- It excels at modeling complex, congested networks with severe queue spillback where traditional macroscopic formulas fail.