Source-linked AI summary
openACC. An open database of car-following experiments to study the properties of commercial ACC systems
Michail Makridis, Konstantinos Mattas, Aikaterini Anesiadou, Biagio Ciuffo
TL;DR
Commercial ACC systems are widely deployed, but their proprietary operation and cross-manufacturer differences remain insufficiently documented. The paper presents openACC, an evolving open database of car-following experiments involving 16 vehicles, including 11 commercial ACC-equipped vehicles, and reports preliminary behavioral findings. The dataset is intended to support empirical study of ACC properties and their possible traffic-flow impacts.
Problem
Commercial ACC operation is largely proprietary, and differences among manufacturers’ systems remain insufficiently documented for studying deployment impacts.
Method
The paper publishes and summarizes openACC, an evolving database built from three car-following campaigns involving commercial ACC systems under public-road and proving-ground conditions.
Results
ACC systems produce smooth acceleration under stable conditions, show string instability under leader speed perturbations, and use headways around 1 s to 2.5 s.
Takeaways & Limitations
openACC enables researchers to observe heterogeneous commercial automation controllers under real-world conditions and develop more precise behavioral and traffic-simulation tools.
Takeaways & Limitations
The database remains an evolving first release, and public-road experiments face disturbances from surrounding vehicles, road geometry, and other obstacles.
Abstract
from arXiv · showhide
Commercial Adaptive Cruise Control (ACC) systems are increasingly available as standard options in modern vehicles. At the same time, still little information is openly available on how these systems actually operate and how different is their behavior, depending on the vehicle manufacturer or model.T o reduce this gap, the present paper summarizes the main features of the openACC, an open-access database of different car-following experiments involving a total of 16 vehicles, 11 of which equipped with state-of-the-art commercial ACC systems. As more test campaigns will be carried out by the authors, OpenACC will evolve accordingly. The activity is performed within the framework of the openData policy of the European Commission Joint Research Centre with the objective to engage the whole scientific community towards a better understanding of the properties of ACC vehicles in view of anticipating their possible impacts on traffic flow and prevent possible problems connected to their widespread. A first preliminary analysis on the properties of the 11 ACC systems is conducted in order to showcase the different research topics that can be studied within this open science initiative.
1 INTRODUCTION
Commercial ACC systems remain difficult to study because their operation is largely proprietary and differences across manufacturers are insufficiently documented. The paper introduces openACC as an open database of car-following experiments designed to support empirical study of ACC behavior and traffic impacts.
- Research gap: Commercial ACC operation remains almost a black box, while differences among manufacturers’ systems are not thoroughly studied.These gaps force researchers to make assumptions when assessing the effects of widespread ACC deployment on public networks.
- Motivation: Experimental campaigns can provide more reliable conclusions but require substantial resources and are constrained by geography and organization.The paper distinguishes area-level traffic observation from vehicle-focused experiments on advanced systems.
- Research gap: Existing traffic datasets capture trajectories and traffic phenomena but generally lack vehicle technical specifications needed to study in-vehicle technologies.NGSIM, pNEUMA, and highD provide valuable trajectory data, yet do not identify the relevant vehicle systems.
- Contribution: openACC makes data from different car-following campaigns publicly available, covering 16 vehicles, including 11 with state-of-the-art commercial ACC systems.The database is intended to grow as additional campaigns are conducted and data are extracted.
- Contribution: The first campaigns examine ACC behavior in two- and three-vehicle, then five-vehicle, car-platoon tests.The larger follow-up campaign investigated response time, time headway, and possible string-stability issues after discomfort was observed during speed perturbations.
2 Experimental campaigns
The first openACC release combines three car-following campaigns on Italian public freeways and the AstaZero proving ground. Their designs progress from exploratory real-world tests to a protected, more systematic study of commercial ACC behavior.
- Campaign overview: Three campaigns comprise the first openACC release, including public-freeway tests with 2, 3, and 5 ACC-equipped vehicles in car-platoon formation.The first two campaigns were conducted on public roads, where surrounding traffic and road geometry complicated platoon preservation.
- Campaign 1: Campaign 1 used two days of testing from Ispra to Cherasco with two or three vehicles, scheduled outside peak hours to reduce disturbances.The leader drove manually with occasional realistic speed perturbations, while followers used ACC whenever possible.
- Campaign 2: Campaign 2 used five vehicles on a three-day Ispra–Vicolungo route to investigate lag-related discomfort and suspected ACC instability from the first campaign.The longer platoon was intended to provide more insight into response behavior under leader accelerations and decelerations.
- Experimental challenges: Real-world tests encountered cut-ins, toll-related platoon inconsistency, low-quality GNSS in tunnels, and overtaking around slow trucks.These conditions motivated manual inspection and later development of more sophisticated filtering.
- Campaign 3: Campaign 3 involved five different high-end vehicle models from four makes over two days on the 5.7-km Rural road at AstaZero.Trajectory data were acquired with the RT-Range S multiple-target ADAS system using differential GNSS.
- Campaign 3: The proving-ground campaign used constant-speed platoons and target-speed perturbations from equilibrium to enable more systematic ACC analysis in a protected environment.The leading vehicle was kept consistent across tests and operated with ACC enabled.
3 Data post-processing
The database combines trajectory and speed measurements from two acquisition systems, supplements them with available vehicle data, and distributes processed campaign data in a simple CSV structure.
- Data acquisition: On-road campaigns used one Ublox M8 device per vehicle, while the test-track campaign used an OXTS inertial-navigation system.The Ublox devices included accelerometers, gyroscopes, and GNSS receivers with reported average horizontal accuracy below 50 cm.
- Data acquisition: Available OBD or CAN data were incomplete and often insufficient for reconstructing accurate trajectory or speed series.When available, these signals supplied accelerator-pedal or driver-assistance information and helped detect ACC activation.
- Data correction: Bumper-to-bumper distances were corrected using antenna-to-bumper measurements for leaders and followers.The correction subtracts the leader’s antenna-back-bumper and follower’s antenna-front-bumper distances.
- Data format: The public dataset is organized into three campaign folders containing CSV car-platoon series, vehicle specifications, and experimental-design descriptions.The CSV structure is documented through the database’s listed columns.
4 Preliminary Results
The openACC preliminary results use heterogeneous car-following experiments to examine commercial ACC behavior, including dynamics, headways, response times, and platoon effects. The findings show manufacturer-dependent behavior, non-instantaneous responses, and distinct differences between ACC-controlled and human-driven vehicles.
- Research applications: The dataset supports analysis of string instability, traffic-flow effects, safety-related spacing, energy demand, and similarities or differences between ACC systems and human drivers.The authors present these as research topics enabled by openACC rather than as fully resolved conclusions.
- Vehicle dynamics and headways: ACC vehicles maintain more consistent speeds and time headways than human-driven vehicles, whose spacing and speed patterns vary more with individual driving behavior.ACC systems target specific headway values, whereas human-driven vehicles show more uniform or random headway distributions and greater speed variability.
- ACC response time: Approximately 2s response times were estimated for two ACC controllers, substantially exceeding instantaneous-response assumptions commonly used in traffic studies.The correlation-based estimates had very high confidence and referred to normal, non-critical driving conditions.
- Vehicle dynamics and headways: ACC systems produce low acceleration variation in stable conditions but sharper acceleration and deceleration responses when the leader introduces perturbations.These sharper responses appear in the acceleration distributions and can become more pronounced upstream in a platoon.
5 Discussion and Conclusions
The openACC database compiles car-following experiments with commercial ACC vehicles and enables preliminary study of their traffic-flow properties. The reported findings contrast ACC behavior with substantial variability among human drivers while emphasizing the need for further validation and broader data collection.
- 11 vehicles equipped with commercial ACC systems form the initial open-access openACC database for studying car-following behavior.The database will expand as additional test campaigns are conducted.
- The dataset supports observing commercial automation controllers under real-world conditions and developing more precise behavioral and traffic-simulation models.The preliminary examples address vehicle behavior, in-vehicle technologies, and simulation-domain tools.
- More structured and systematic analysis is needed to validate and confirm the preliminary findings.Future work includes longer car-platoons, additional ADAS measurements, and deeper investigation of automation effects on traffic and energy demand.
- Commercial ACC systems show smooth acceleration and deceleration, non-negligible response times, headways around 1s to 2.5s, and string instability under small time-headway settings.The reported response times are comparable to, or not higher than, those of attentive human drivers.
- ACC driving behavior is more homogeneous than that of human drivers, whose time headway, acceleration, deceleration, and response-time strategies vary substantially.Human drivers can anticipate upstream perturbations and absorb traffic oscillations, whereas the ACC findings motivate attention to string instability.
A. Time headway distributions from the campaign N.1
Campaign N.1 compares time-headway distributions for human drivers and ACC followers in a car-platoon. Figure 14 presents the distributions for all vehicles involved in the campaign.
- Time-headway distributions are shown for human-driver and ACC-system followers in the campaign N.1 car-platoon.The comparison concerns the two vehicles driving as followers during the first campaign.
- Figure 14 presents time-headway distributions for all vehicles involved in campaign N.1.
B. Minimum response time estimates
The paper reports minimum response-time estimates for the 11 ACC systems under study. These values are reference estimates rather than deterministic controller characteristics.
- Minimum response-time estimates are reported for all 11 ACC systems under study.The estimates represent minimum values under freeway, non-critical driving conditions.
- The controllers’ response-time values are estimates indicating order of magnitude because controller functionality is not deterministic.
C. Acceleration distributions with differential GNSS
Figure 15 compares acceleration and deceleration distributions by vehicle and driving mode. The campaign’s differential-GNSS test-track data are shown before and after moving-average filtering.
- The test-track campaign’s raw acceleration measurements are compared with data after applying a moving-average filter.Unlike the on-road tests, this campaign did not require post-filtering to produce accelerations within realistic boundaries.
- Figure 15 shows acceleration and deceleration distributions per vehicle and driving mode.A prefix “(L)” identifies vehicles used as leaders of the car-platoon.
D. String instability of ACC systems
Figure 16 compares car-platoon behavior under ACC and human driving across campaigns N.2 and N.3. The examples show evident string instability with ACC, while it is not obvious in the all-human platoon example.
- Campaign N.2: In campaign N.2, a five-vehicle platoon with four ACC-equipped followers exhibits obvious string instability.The corresponding speed series is shown in Fig. 16a.
- Campaign N.2: The all-human-driven platoon from campaign N.2 does not show obvious string instability in the corresponding example.This comparison is presented in Fig. 16b.
- Campaign N.3: Figures 16c and 16d provide more obvious partial-trajectory examples from campaign N.3 for comparing ACC-system and human-driver behavior.The passage introduces these figures as a clearer comparison, but the supplied text does not state the specific outcome beyond that contrast.
E. Indicative template of the information file
The openACC information file organizes campaign-level documentation, vehicle details, equipment, processing, trip annotations, and CSV column descriptions. Its fields cover both experimental metadata and the structure of the released data.
- File contents: Each openACC database folder contains vehicle specifications and a short description of the associated experimental campaign.The information file also serves as a sample of the database’s structure.
- File contents: The experiment description records the campaign’s main objectives and the number of included trips.This field provides the primary context for the experiment.
- File contents: Vehicle specifications identify the vehicles and report their specifications, while a separate field records the number of vehicles in each campaign.These fields document both vehicle characteristics and campaign composition.
- File contents: The equipment field identifies the data-acquisition equipment used during the campaign.It is listed as a dedicated component of the information file.
- File contents: Data processing documents the technique used to achieve a 10 Hz frequency, while trip comments and columns info describe trip data and CSV fields.Together, these entries explain processing, trip-specific annotations, and the released file columns.