2.2 Microscopic Traffic Characteristics & Car-Following Models

2.2 Microscopic Traffic Characteristics & Car-Following Models

PTOE Exam Focus: Microscopic traffic flow bridges individual driver-vehicle behavior with system-wide stream dynamics. Key exam proficiencies include calculating headway and gap probabilities using the negative exponential and shifted exponential distributions, applying the General Motors (GHR) stimulus-response car-following equation, deriving macroscopic stream equations from GHR sensitivity exponents, and evaluating platoon string stability criteria ($C = \alpha \Delta t$).

---

1. Microscopic Traffic Stream Variables & Dualities

While macroscopic parameters describe the aggregate stream, microscopic parameters govern the physical spacing and temporal intervals between consecutive vehicles ($n$ and $n+1$):

 Vehicle n+1 (Follower)                           Vehicle n (Leader)
[====Front-Bumper====]                          [====Front-Bumper====]
         |--------------- Space Gap (g_s) ---------------|      |
         |<------------ Space Headway (h_s) ------------>|--L_v-|

Definitions & Relationships:

1. Time Headway ($h_t$ or $h$): The elapsed time between the arrival of corresponding points (typically front bumpers) of two consecutive vehicles at a specific cross-section, measured in seconds/vehicle: $\bar{h}_t = \frac{3600}{q}$ Where $q$ is flow rate in veh/hr/ln and $\bar{h}_t$ is mean time headway.

2. Time Gap ($g_t$): The elapsed time between the rear bumper of the leading vehicle and the front bumper of the following vehicle: $g_t = h_t - \frac{L_v}{u}$ Where $L_v$ is vehicle length (ft) and $u$ is vehicle speed (ft/s).

3. Space Headway ($h_s$ or $s$): The physical distance between corresponding points (front bumpers) of two consecutive vehicles at a given instant, measured in feet/vehicle or meters/vehicle: $\bar{h}_s = \frac{5280}{k}$ Where $k$ is density in veh/mi/ln and $\bar{h}_s$ is mean space headway.

4. Space Gap ($g_s$ or Clearance): The clear physical distance between the rear bumper of the lead vehicle and the front bumper of the following vehicle: $g_s = h_s - L_v$

Fundamental Micro-Macro Duality:

Combining the mean headway definitions yields the exact bridge between micro and macro streams: $\bar{u}_s = \frac{\bar{h}_s}{\bar{h}_t} \times \left(\frac{3600\text{ s/hr}}{5280\text{ ft/mi}}\right) = \frac{q}{k}$

---

2. Statistical Distributions of Headways

Vehicle arrival headways are stochastic. Selection of the appropriate statistical distribution depends directly on the traffic volume and degree of vehicle interaction.

A. Negative Exponential Distribution (Random Arrivals / Poisson Process)

When traffic flow is light ($q < 400\text{ veh/hr/ln}$), vehicles arrive independently without mutual interference. The number of vehicle arrivals in time interval $t$ follows a discrete Poisson distribution: $P(n\text{ arrivals in interval } t) = \frac{(\lambda t)^n e^{-\lambda t}}{n!}$ Where $\lambda = \frac{q}{3600}$ is the mean arrival rate in vehicles per second.

The probability that time headway $h$ is greater than or equal to time $t$ corresponds to the probability of zero arrivals ($n=0$) during interval $t$, yielding the Negative Exponential Distribution: $P(h \ge t) = e^{-\lambda t} = \exp\left(-\frac{q t}{3600}\right)$ $P(h < t) = 1 - e^{-\lambda t}$

• Probability Density Function (PDF): $f(t) = \lambda e^{-\lambda t}$ for $t \ge 0$.
• Mean: $E[h] = \frac{1}{\lambda} = \bar{h}_t$.
• Variance: $\operatorname{Var}(h) = \frac{1}{\lambda^2} = \bar{h}_t^2$.

B. Shifted Exponential Distribution

The pure exponential distribution unrealistically assigns the highest probability density to $t = 0$. In reality, physical vehicle length and minimum safe following buffer impose a lower bound on headway (minimum headway $\tau \approx 0.5\text{--}1.2\text{ s}$): $P(h \ge t) = e^{-\lambda' (t - \tau)} \quad \text{for } t \ge \tau$ Where $\lambda' = \frac{\lambda}{1 - \lambda \tau} = \frac{q}{3600 - q\tau}$.

C. Composite & Bunched Headway Models

At moderate-to-heavy flow rates ($q > 600\text{ veh/hr/ln}$), vehicles group into platoons:

• Cowan M3/M4 Models: Divide the stream into a proportion $\alpha$ of free vehicles (exponential headways) and $(1-\alpha)$ of tracking/bunched vehicles (constrained at headway $\tau$).
• Lognormal & Gamma Distributions: Fit empirical headway data where the mode shifts away from zero to typical following headways of $1.5\text{--}2.2\text{ seconds}$.

---

3. Car-Following Formulations (Microscopic Stimulus-Response Theory)

Car-following models mathematically govern how a trailing vehicle ($n+1$) accelerates or decelerates in response to the motion of the preceding vehicle ($n$) within a single traffic lane.

$\text{Response}(t + \Delta t) = \text{Sensitivity} \times \text{Stimulus}(t)$ Where $\Delta t$ is the driver perception-reaction time (typically $1.0\text{--}1.5\text{ s}$).

A. Pipes' Model (California Code Rule-of-Thumb)

Louis A. Pipes (1953) formalized the traditional DMV safe driving rule: "allow at least one car length of space for every 10 mph of speed." $s_{\min} = L_v + \left(\frac{u}{10}\right) L_v = L_v \left(1 + \frac{u}{10}\right)$ Assuming standard vehicle length $L_v = 20\text{ ft}$, spacing in feet as a function of speed $u$ (mph) becomes: $s_{\min} = 20 + 2u$ Expressed in terms of time headway (since $u\text{ in ft/s} = 1.467 u\text{ in mph}$): $h_t = \frac{s_{\min}}{1.467 u} = \frac{20 + 2u}{1.467 u} = 1.363 + \frac{13.63}{u}$ At high speeds, time headway asymptotically approaches a constant safety buffer of $1.36\text{ seconds}$.

B. General Motors (Gazis-Herman-Rothery / GHR) Non-Linear Model

Developed between 1958 and 1961 at the GM Research Laboratories, the GHR model represents the generalized non-linear stimulus-response formulation: $\ddot{x}_{n+1}(t + \Delta t) = \alpha \frac{[\dot{x}_{n+1}(t)]^m}{[x_n(t) - x_{n+1}(t)]^l} \left[\dot{x}_n(t) - \dot{x}_{n+1}(t)\right]$ Where:

• $\ddot{x}_{n+1}(t + \Delta t)$ = acceleration/deceleration response of follower at time $t + \Delta t$
• $[\dot{x}n(t) - \dot{x}{n+1}(t)] = \Delta v(t)$ = relative speed stimulus at time $t$
• $[x_n(t) - x_{n+1}(t)] = s(t)$ = space headway / distance separation
• $\dot{x}_{n+1}(t)$ = speed of follower
• $\alpha$ = sensitivity scaling coefficient
• $m$ = speed sensitivity exponent
• $l$ = distance/spacing sensitivity exponent

---

4. The Micro-Macro Mathematical Bridge

A central theoretical achievement of traffic flow theory is proving that macroscopic continuum models are direct mathematical integrals of the microscopic GHR car-following model under steady-state conditions ($s = 1/k$, $u = \dot{x}$):

$\frac{du}{dk} = -\frac{\alpha}{k^{2-l} u^{-m}}$

Exam Rule: If an exam question specifies GHR parameters $m=0, l=2$, the macroscopic stream follows Greenshields; if $m=0, l=1$, it follows Greenberg; if $m=1, l=1$, it follows Underwood; and if $m=1, l=2$, it follows the Northwestern (Drake) bell-curve model.

---

5. Platoon Stability & Disturbance Propagation

Platoon stability analysis evaluates how a speed perturbation (such as tapping the brake) introduced by a lead vehicle propagates backward through a following platoon.

Stability Classifications:

1. Local Stability: Concerns the behavior of a single follower relative to its immediate leader. The perturbation dampens if the follower's response does not overshoot.
2. Asymptotic / String Stability: Concerns the propagation of the disturbance through an entire line of $N$ vehicles. If the disturbance amplifies with each successive follower, the platoon is asymptotically unstable, creating a stop-and-go shockwave and high risk of rear-end collisions.

The Stability Parameter ($C$):

For linear car-following ($m=0, l=0$), stability is dictated by the dimensionless product of sensitivity $\alpha$ and perception-reaction time $\Delta t$: $C = \alpha \Delta t$

Two different thresholds govern the two stability questions, and confusing them is a classic exam trap.

Local stability (how one follower responds to its own leader):

• $C < \frac{1}{e} \approx 0.368$: The follower's spacing error decays monotonically — no oscillation at all.
• $\frac{1}{e} < C < \frac{\pi}{2} \approx 1.57$: The response oscillates, but the amplitude decays (damped oscillation).
• $C = \frac{\pi}{2} \approx 1.57$: Neutral stability — oscillation of constant, undamped amplitude.
• $C > \frac{\pi}{2} \approx 1.57$: Locally unstable — the follower's oscillation grows.

Asymptotic / string stability (how the disturbance propagates along $N$ vehicles):

• $C < 0.50$: String stable. The perturbation attenuates with each successive follower upstream.
• $C > 0.50$: String unstable. Oscillations amplify down the platoon , and the magnitude of deceleration grows until trailing vehicles must execute emergency stops ($a = a_{\max}$), causing phantom traffic jams or multi-vehicle crashes.

Exam Rule: $C = 0.40$ is string stable ($C < 0.50$) but not locally non-oscillatory ($C > 1/e = 0.368$) — the disturbance dies out along the platoon while still producing damped oscillation in each individual follower.