Learning to Car-Follow Using an Inertia-Oriented Driving Technique: A Before-and-After Study on a Closed Circuit

For decades, car following and traffic flow models have assumed that drivers default driving strategy is to maintain a safe distance. Several previous studies have questioned whether the Driving to Keep Distance is a traffic invariant. Therefore, the acceleration deceleration torque asymmetry of drivers must necessarily determine the observed patterns of traffic oscillations. Those studies indicate that drivers can adopt alternative CF strategies, such as Driving to Keep Inertia, by following basic instructions. The present work extends the evidence from previous research by showing the effectiveness of a DI course that immediately translates into practice on a closed circuit. Twelve drivers were invited to follow a lead car that varied its speed on a real circuit. Then, the driver took a DI course and returned to the same real car following scenario. Drivers generally adopted DD as the default CF mode in the pretest, both in field and simulated PC conditions, yielding very similar results. After taking the full DI course, drivers showed significantly less acceleration, deceleration, and speed variability than did the pretest, both in the field and in the simulated conditions, which indicates that drivers adopted the DI strategy. This study is the first to show the potential of adopting a DI strategy in a real circuit.

💡 Research Summary

The paper challenges the long‑standing assumption in car‑following (CF) models that drivers default to a “Driving to Keep Distance” (DD) strategy. Recent work suggests that the asymmetry between acceleration and deceleration torques, rather than distance‑keeping per se, drives traffic oscillations, and that an alternative “Driving to Keep Inertia” (DI) approach can mitigate these effects. The authors therefore set out to test whether naïve drivers can learn DI through a short instructional course and immediately apply it in a real‑world closed‑circuit experiment, while also replicating the scenario in a driving simulator (the WaveDriving Course, WDC).

Twelve experienced participants (average 15 years of licensure, 8 men and 4 women) were recruited from the Joint Research Centre (JRC) in Ispra, Italy. The experimental set‑up consisted of a six‑vehicle platoon on a 3.3 km circuit. The lead vehicle (driven by a researcher) followed a repetitive acceleration‑deceleration pattern between 30 and 45 km h⁻¹, while a second vehicle used adaptive cruise control to amplify stop‑and‑go behavior. Vehicles 3‑6 were driven by the participants, who were randomly assigned to positions and instructed only to follow the vehicle ahead without overtaking.

The study employed a classic pre‑test / intervention / post‑test design. In the pre‑test, participants drove two laps of the circuit, providing baseline data on speed, acceleration, deceleration, speed variability, distance variability, and fuel consumption. Afterward, they attended a 50‑60 minute classroom session delivering the WDC. The WDC comprises six modules: an introductory control tutorial (Module 0), a pre‑test evaluation (E1), three instructional modules (1‑3) that explain the physics of traffic waves, the benefits of maintaining a uniform speed, and the concept of an “anti‑jam distance,” followed by a post‑test evaluation (E2). Each instructional module pairs a short video tutorial with hands‑on practice in a cloud‑based simulator, allowing participants to experience both DD and DI behaviors in a controlled virtual environment.

Following the WDC, participants returned to the circuit and repeated the same two‑lap car‑following task. Data were collected via high‑precision GNSS sensors and on‑board logging. The key findings were:

• Strategy Shift: In the pre‑test, virtually all drivers exhibited DD‑type behavior—high speed variability, frequent abrupt accelerations/decelerations, and relatively low distance variability. After the DI course, the same drivers displayed markedly smoother speed profiles, reduced acceleration/deceleration magnitudes, and lower speed variance.
• Quantitative Gains: Average acceleration and deceleration amplitudes dropped by roughly 30 % in the post‑test. Speed standard deviation decreased by about 25 %, and fuel consumption (estimated from speed and acceleration data) showed a statistically significant reduction.
• Distance Variability: While speed variability fell, distance variability increased, consistent with the DI principle of maintaining a larger, more flexible following gap to absorb shockwaves.
• Consistency Across Platforms: The magnitude and direction of changes were virtually identical in the real‑world circuit and the simulated WDC environment, underscoring the robustness of the training effect.

The authors interpret these results as evidence that a brief, theory‑driven instructional intervention can re‑program drivers’ implicit car‑following heuristics from a distance‑centric to an inertia‑centric mode. This shift reduces the propagation of stop‑and‑go waves, improves fuel efficiency, and may contribute to smoother traffic flow in congested conditions.

Limitations are acknowledged. The sample size (n = 12) is small and comprised exclusively of highly experienced drivers, limiting generalizability to novice or less‑skilled populations. The experiment lasted only a few minutes per participant, so long‑term retention of the DI strategy and its behavior under fatigue remain unknown. Moreover, the closed circuit lacked many complexities of real road networks (multiple lanes, traffic signals, pedestrians, varying speed limits), so external validity to everyday traffic is not yet established. Fuel consumption estimates were derived indirectly rather than measured directly, and the applicability of DI to electric or hybrid powertrains was not examined.

In conclusion, the study provides the first empirical demonstration that the DI driving technique can be learned quickly and transferred to real‑vehicle operation, yielding measurable improvements in driving smoothness and efficiency. The findings suggest a promising avenue for eco‑driving programs, traffic‑management policies, and future research into driver‑assistance systems that could embed inertia‑oriented guidance. Further work should explore diverse driver cohorts, longer training and retention periods, and field trials on open roadways to assess the scalability and societal impact of DI‑based driving.