Search

• This paper has characterized the empirical car-following (CF) behaviors of a commercial ACC system and explained the underlying mechanism(s). • The ACC behaviors (i.e., response time, oscillation growth, and overshoot) largely depend on the ACC headway, speed level, and leader stimulus. • In a platoon, the oscillation growth progressively changes from one ACC to another, implying oscillation escalation and potential safety hazards. Emerging automated vehicle (AV) technologies are increasingly being deployed around the world and it is only a matter of time before the transportation landscape changes dramatically. Unfortunately, those changes cannot be well predicted due to the lack of empirical data. But adaptive cruise control (ACC) vehicles are common in the market and can be used to fill this gap. In this paper, we aim to characterize the empirical car-following behaviors of a commercial ACC system and understand how ACC behaves in different conditions and the underlying impact mechanism. It is found that for a single ACC: (i) the ACC response time is comparable to human drivers but much larger than the ACC controller time gap and it exhibits small variance, (ii) the ACC response can amplify or dampen an oscillation, (iii) after the oscillation, the stabilization process can exhibit overshooting or undershooting, and (iv) these CF behaviors depend largely on the ACC headway setting, speed level, and leader stimulus, which produce the impacts directly and/or indirectly through the mediation of earlier ACC behaviors. For a three-vehicle platoon, our main finding is that the change from one ACC vehicle to the next is progressive for oscillation growth, and regressive for deceleration, acceleration, and overshooting. This implies that in long platoons, oscillation amplitude tends to exacerbate very quickly, which forces ACC vehicles further upstream to apply very strong braking followed by a strong acceleration. This can cause significant overshooting and safety hazards. [ABSTRACT FROM AUTHOR]

• Investigate impacts of heterogeneity in driving behavior on traffic dynamics. • Unifying framework to analyze RV and CAV behavior in mixed traffic. • Mechanism for heterogenous acceleration and CF leading to void creation. • Simulation for compounded effects of heterogeneity in mixed traffic platoons. This paper elucidates the impacts of vehicle heterogeneity on traffic dynamics and throughput of mixed traffic consisting of connected automated vehicles (CAVs) and regular vehicles (RVs). The main premise is that the heterogeneity in preferred acceleration rate, desired speed, and car-following (CF) behavior (e.g., reaction pattern and sensitivity to a traffic disturbance) will change traffic properties in ways that can undermine traffic flow throughput. This paper first decomposes the mechanism into two elements – one driven by acceleration and one by time-varying CF response to disturbances – and then investigates their compounded effect. This paper also provides unifying frameworks to analyze the behavior of RVs and CAVs to facilitate analytical investigations. The results reveal how heterogeneous acceleration and CF behavior may create persistent voids and diminish traffic throughput. Integrating all the elements, throughput reduction is quantified via numerical simulations. [ABSTRACT FROM AUTHOR]

This paper examines CACC truck platooning on uphill grades. It was found that the design of CT policy should consider the effects of low crawl speeds on significant upgrades. Three simple solutions, which have different impacts on traffic flow efficiency, are proposed. Furthermore, truck platoons, controlled by a state-of-the-art CACC model, become asymptotically unstable beyond some critical grade. The errors are permanent, suggesting that trucks fail to re-engage after the upgrade. This occurs by complex interactions between the CACC control and the bounded acceleration capabilities of trucks. New control concepts are developed to complement the existing control model and achieve asymptotic (and string) stability. The instability mechanisms and new control concepts are not specific to the control model used. [ABSTRACT FROM AUTHOR]

This paper investigates the mechanisms of how spatially distributed lane changes (LCs) interact and contribute to “capacity-drop” at three types of extended bottlenecks: merge, diverge, and weave. A hybrid approach is used to study the problem: analytical approach to capture the behavior of merging and diverging LCs and numerical simulations to quantify capacity-drop considering various geometric configurations of extended bottlenecks. This study focuses on the impact of LC vehicles’ bounded acceleration on “void” (wasted space) creation in traffic streams when they insert/desert at a lower speed, and interactions among multiple voids. We found that (1) LCs closer to the downstream end of bottlenecks are more likely to create persisting voids and contribute to capacity-drop. (2) For weave bottlenecks, capacity-drop is governed by two counteracting effects of LCs: persisting voids and utilization of vacancies created by diverging vehicles; (3) the more balanced the merging and diverging flows, the lower the capacity-drop; and (4) capacity-drop is minimum if merging LCs occur downstream of diverging LCs, and maximum in the opposite alignment. [ABSTRACT FROM AUTHOR]

• We develop a generic strategy for any CAV-HDV topology in mixed traffic to efficiently stabilize the traffic oscillations. • We propose a novel car-following structure, 'CAV-HDVs-CAV' which can better capture HDVs' stochasticity by aggregation. • Our method marries the merits of traffic flow theory, DRL, and network control theory organically for CAV control, which shows a very promising control performance. This paper proposes an innovative distributed longitudinal control strategy for connected automated vehicles (CAVs) in the mixed traffic environment of CAV and human-driven vehicles (HDVs), incorporating high-dimensional platoon information. For mixed traffic, the traditional CAV control method focuses on microscopic trajectory information, which may not be efficient in handling the HDV stochasticity (e.g., long reaction time; various driving styles) and mixed traffic heterogeneities. Different from traditional methods, our method, for the first time, characterizes consecutive HDVs as a whole (i.e., AHDV) to reduce the HDV stochasticity and utilize its macroscopic features to control the following CAVs. The new control strategy takes advantage of platoon information to anticipate the disturbances and traffic features induced downstream under mixed traffic scenarios and greatly outperforms the traditional methods. In particular, the control algorithm is based on deep reinforcement learning (DRL) to fulfill car-following control efficiency and further address the stochasticity for the aggregated car following behavior by embedding it in the training environment. To better utilize the macroscopic traffic features, a general platoon of mixed traffic is categorized as a CAV-HDVs-CAV pattern and described by corresponding DRL states. The macroscopic traffic flow properties are built upon the Newell car-following model to capture the characteristics of aggregated HDVs' joint behaviors. Simulated experiments are conducted to validate our proposed strategy. The results demonstrate that the proposed control method has outstanding performances in terms of oscillation dampening, eco-driving, and generalization capability. [ABSTRACT FROM AUTHOR]

This paper provides formulations of traffic operational capacity in mixed traffic, consisting of automated vehicles (AVs) and regular vehicles, when traffic is in equilibrium. The capacity formulations take into account (1) AV penetration rate, (2) micro/mesoscopic characteristics of regular and automated vehicles (e.g., platoon size, spacing characteristics), and (3) different lane policies to accommodate AVs such as exclusive AV and/or RV lanes and mixed-use lanes. A general formulation is developed to determine the valid domains of different lane policies and more generally, AV distributions across lanes with respect to demand, as well as optimal solutions to accommodate AVs. [ABSTRACT FROM AUTHOR]

The connected vehicle (CV) technology is applied to develop VSL strategies to improve bottleneck discharge rates and reduce system delays. Three VSL control strategies are developed with different levels of complexity and capabilities to enhance traffic stability using: (i) only one CV (per lane) (Strategy 1), (ii) one CV (per lane) coupled with variable message signs (Strategy 2), and (iii) multiple CVs (Strategy 3). We further develop adaptive schemes for the three strategies to remedy potential control failures in real time. These strategies are designed to accommodate different queue detection schemes (by CVs or different sensors) and CV penetration rates. Finally, probability of control failure is formulated for each strategy based on the stochastic features of traffic instability to develop a general framework to (i) estimate expected delay savings, (ii) assess the stability of different VSL control strategies, and (iii) determine optimal control speeds under uncertainty. Compared to VMS-only strategies, the CV-based strategies can effectively impose dynamic control over continuous time and space, enabling (i) faster queue clearance around a bottleneck, (ii) less restrictive control with higher control speed (thus smoother transition), and (iii) simpler control via only one or a small number of CVs. [ABSTRACT FROM AUTHOR]

Variable speed limit (VSL) schemes are developed based on the Kinematic Wave theory to increase discharge rates at severe freeway bottlenecks induced by non-recurrent road events such as incidents or work zones while smoothing speed transition. The main control principle is to restrict upstream demand (in free-flow) progressively to achieve three important objectives: (i) to provide gradual speed transition at the tail of an event-induced queue, (ii) to clear the queue around the bottleneck, and (iii) to discharge traffic at the stable maximum flow that can be sustained at the bottleneck without breakdown. These control objectives are accomplished without imposing overly restrictive speed limits. We further provide remedies for (a) underutilized bottleneck capacity due to underestimated stable maximum flow and (b) a re-emergent queue at the bottleneck due to an overestimated stable maximum flow. We analytically formulate the reductions in total delay in terms of control parameters to provide an insight into the system performance and sensitivity. The results from the parameter analysis suggest that significant delay savings can be realized with the proposed VSL control strategies. [ABSTRACT FROM AUTHOR]

New variable speed limit (VSL) schemes are developed based on the Kinematic Wave theory to increase freeway bottleneck discharge rates under two scenarios: (i) steady queue and (ii) oscillatory queue. The key principle is to impose VSL control some distance upstream of a bottleneck to starve the inflow to the bottleneck and dissipate the queue. Once the queue near the bottleneck vanishes, another less restrictive VSL is imposed upstream to (i) resolve the heavy queue generated by the first VSL and (ii) regulate the inflow to the bottleneck to sustain the stable maximum bottleneck discharge rate and prevent traffic breakdown. Several strategies are developed for each scenario ranging from the simplest strategy that maximizes the delay saving to more sophisticated strategies for upstream queue management. An analysis of the model parameters suggests that a wide range of the first speed limit (to clear the queue around the bottleneck) can be imposed to realize significant delay savings. [ABSTRACT FROM AUTHOR]

Experimental measurements on commercial adaptive cruise control (ACC) vehicles are becoming increasingly available from around the world, providing an unprecedented opportunity to study the traffic flow characteristics that arise from this technology. This paper adds new experimental evidence to this knowledge base and presents a comprehensive empirical study on the ACC equilibrium behaviors via the resulting fundamental diagrams. We find that, like human-driven vehicles, ACC systems display a linear equilibrium spacing-speed relationship (within the range of available data) but the key parameters of these relationships can differ significantly from human-driven traffic depending on input settings. At the minimum ACC headway setting, equilibrium capacities in excess of 3500 vehicles per hour are observed, together with an extremely fast equilibrium wave speed of 100 kilometers per hour. These fast waves are unfamiliar to human drivers, and may pose a safety risk. The results also suggest that ACC jam spacing can be much larger than in human traffic, which reduces the network storage capacity. • A comprehensive study on ACC fundamental diagrams based on worldwide experiments. • ACC has a linear speed-spacing relationship in the available data range. • ACC can produce high capacities excessing 3500 vph with fast congested wave speed. • ACC jam spacing is much larger than human, which reduces network storage capacity. [ABSTRACT FROM AUTHOR]

Highlights: [•] Traffic hysteresis arises by variable driver characteristics within each driver. [•] We explain the mechanism of periodicity of traffic oscillations. [•] Aggressive driver behavior leads to hysteresis loops that induce the upstream propagation of oscillations. [•] Larger hysteresis loops induce larger oscillation growth. [•] Variable driver characteristics induce capacity drop. [ABSTRACT FROM AUTHOR]

Abstract: This paper studies traffic hysteresis arising in traffic oscillations from a behavioral perspective. It is found that the occurrence and type of traffic hysteresis is closely correlated with driver behavior when experiencing traffic oscillations and with the time driver reaction begins relative to the starting deceleration wave. Statistical results suggest that driver behavior is different depending on its position along the oscillation. This suggests that different car-following models should be used inside the different stages of an oscillation in order to replicate realistic congestion features. [Copyright &y& Elsevier]

Abstract: This paper presents a behavioral car-following model based on empirical trajectory data that is able to reproduce the spontaneous formation and ensuing propagation of stop-and-go waves in congested traffic. By analyzing individual drivers’ car-following behavior throughout oscillation cycles it is found that this behavior is consistent across drivers and can be captured by a simple model. The statistical analysis of the model’s parameters reveals that there is a strong correlation between driver behavior before and during the oscillation, and that this correlation should not be ignored if one is interested in microscopic output. If macroscopic outputs are of interest, simulation results indicate that an existing model with fewer parameters can be used instead. This is shown for traffic oscillations caused by rubbernecking as observed in the US 101 NGSIM dataset. The same experiment is used to establish the relationship between rubbernecking behavior and the period of oscillations. [Copyright &y& Elsevier]

• This paper proposes three control strategies for lane-changing in CAV mixed traffic at a merge bottleneck to mitigate voids (thus capacity-drop) and speed disturbances and therefore improve system performance. • In a general control framework, the three strategies are integrated in different combinations exploiting their complementary nature. • The effects of control on capacity-drop are analytically derived and compared with the baseline. This paper proposes three control strategies for lane-changing (LC) at a merge bottleneck to improve bottleneck throughput by mitigating voids and speed disturbances in mixed traffic using connected and automated vehicle (CAV) technologies. Strategy 1 is 'gap closure' control, where an LC vehicle and its follower are controlled to close the void (extra time gap ahead) induced by the LC and prevent a backward-propagating speed disturbance. Strategy 2 is 'batch LC', where a group of LC vehicles are controlled to line up along a kinematic wave to minimize the total voids. Strategy 3 is 'gap redistribution' control, where extra gaps of vehicles are redistributed to periodically create large enough gaps for disturbance-free insertions. In a general control framework, the three strategies are integrated in different combinations exploiting their complementary nature. A numerical analysis shows that certain combinations, such as Strategy 2 and 3, can be very effective for improving bottleneck throughput. The analysis reveals insights on leveraging CAVs to develop traffic management strategies and/or policies and therefore improve system performance. [ABSTRACT FROM AUTHOR]

• Car following behavioral stochasticity has been analyzed based on two sets of field experimental trajectory data. • Wave travel time does not follow a consistent pattern. • Changing rate of wave travel time follows a mean reversion process. • A new stochastic Newell type model is proposed. • Stochastic factors exhibit common feature from different perspectives. This paper analyzes the car following behavioral stochasticity based on two sets of field experimental trajectory data by measuring the wave travel time series τ ˜ n (t) of vehicle n. The analysis shows that (i) No matter the speed of leading vehicle oscillates significantly or slightly, τ ˜ n (t) might change significantly; (ii) A follower's τ ˜ n (t) can vary from run to run even if the leader travels at the same stable speed; (iii) Sometimes, even if the leader's speed fluctuates significantly, the follower can keep a nearly constant value of τ ˜ n (t). The Augmented Dickey-Fuller test indicates that the time series ξ n (t) = d τ ˜ n (t) / d t follows a mean reversion process, no matter the oscillations fully developed or not. Based on the finding, a simple stochastic Newell model is proposed. The concave growth pattern of traffic oscillations has been derived analytically. Furthermore, simulation results demonstrate that the new model well captures both macroscopic characteristic of traffic flow evolution and microscopic characteristic of car following. [ABSTRACT FROM AUTHOR]

• Waymo Open Dataset was comprehensively and systematically processed and assessed. • Consistency analysis, jerk value analysis and trajectory completeness analysis were implemented. • Outliers were removed and noise was filtered. • The impact of data outliers and noise on IDM calibration was assessed. • The processed dataset is publicly available. Recently released Autonomous Vehicle (AV) trajectory datasets can potentially catalyze research progress on AV-oriented traffic flow analysis. This paper aims to comprehensively and systematically process and assess one of the AV-oriented open datasets, i.e., Waymo Open Dataset, with a focus on car following paired trajectories. First, the original dataset has been processed into a user-friendly format which contains all important information related to the behavior of AV and surrounding objects. Second, the data quality has been assessed in terms of internal consistency, jerk values and trajectory completeness. Results show that the extracted trajectories are all incomplete but generally they have better quality than that of Next Generation Simulation program (NGSIM) dataset. Third, the trajectory data has been further enhanced by using an optimization-based outlier removal method and a wavelet denoising method. Additionally, we have tested the impact of data outliers and noise on IDM calibration, and revealed significant differences in parameter values for desired time gap T and maximum acceleration a. [ABSTRACT FROM AUTHOR]

This paper develops a stochastic and unifying framework to examine variability in car-following (CF) dynamics of commercial automated vehicles (AVs) and its direct relation to traffic-level dynamics. The asymmetric behavior (AB) model by Chen et al. (2012a) is extended to accommodate a range of CF behaviors by AVs and compare with the baseline of human-driven vehicles (HDVs). The parameters of the extended AB (EAB) model are calibrated using an adaptive sequential Monte Carlo method for Approximate Bayesian Computation (ABC-ASMC) to stochastically capture various uncertainties including model mismatch resulting from unknown AV CF logic. The estimated posterior distributions of the parameters reveal significant differences in CF behavior (1) between AVs and HDVs, and (2) across AV developers, engine modes, and speed ranges, albeit to a lesser degree. The estimated behavioral patterns and simulation experiments further reveal mixed platoon dynamics in terms of traffic throughout reduction and hysteresis. • Extended Asymmetric Behavioral (EAB) model enables a unifying car-following analysis. • The EAB model is calibrated stochastically via Approximate Bayesian Computation. • Car-following behaviors of automated vehicles exhibit significant heterogeneity. • Car-following dynamics of automated vehicles are directly linked to traffic dynamics. [ABSTRACT FROM AUTHOR]

Abstract: This paper investigates the effects of lane-changing in driver behavior by measuring (i) the induced transient behavior and (ii) the change in driver characteristics, i.e., changes in driver response time and minimum spacing. We find that the transition largely consists of a pre-insertion transition and a relaxation process. These two processes are different but can be reasonably captured with a single model. The findings also suggest that lane-changing induces a regressive effect on driver characteristics: a timid driver (characterized by larger response time and minimum spacing) tends to become less timid and an aggressive driver less aggressive. We offer an extension to Newell’s car-following model to describe this regressive effect and verify it using vehicle trajectory data. [Copyright &y& Elsevier]

Abstract: In this paper we identify the origins of stop-and-go (or slow-and-go) driving and measure microscopic features of their propagations by analyzing vehicle trajectories via Wavelet Transform. Based on 53 oscillation cases analyzed, we find that oscillations can be originated by either lane-changing maneuvers (LCMs) or car-following (CF) behavior. LCMs were predominantly responsible for oscillation formations in the absence of considerable horizontal or vertical curves, whereas oscillations formed spontaneously near roadside work on an uphill segment. Regardless of the trigger, the features of oscillation propagations were similar in terms of propagation speed, oscillation duration, and amplitude. All observed cases initially exhibited a precursor phase, in which slow-and-go motions were localized. Some of them eventually transitioned into a well-developed phase, in which oscillations propagated upstream in queue. LCMs were primarily responsible for the transition, although some transitions occurred without LCMs. Our findings also suggest that an oscillation has a regressive effect on car-following behavior: a deceleration wave of an oscillation affects a timid driver (characterized by larger response time and/or minimum spacing) to become less timid and an aggressive driver less aggressive, although this change may be short-lived. An extended framework of Newell’s CF model is able to describe the regressive effect with two additional parameters with reasonable accuracy, as verified using vehicle trajectory data. [Copyright &y& Elsevier]

Abstract: This paper demonstrates the capabilities of wavelet transform (WT) for analyzing important features related to bottleneck activations and traffic oscillations in congested traffic in a systematic manner. In particular, the analysis of loop detector data from a freeway shows that the use of wavelet-based energy can effectively identify the location of an active bottleneck, the arrival time of the resulting queue at each upstream sensor location, and the start and end of a transition during the onset of a queue. Vehicle trajectories were also analyzed using WT and our analysis shows that the wavelet-based energies of individual vehicles can effectively detect the origins of deceleration waves and shed light on possible triggers (e.g., lane-changing). The spatiotemporal propagations of oscillations identified by tracing wavelet-based energy peaks from vehicle to vehicle enable analysis of oscillation amplitude, duration and intensity. [ABSTRACT FROM AUTHOR]

Currently, model predictive control (MPC) for adaptive cruise control (ACC) systems relies on the prediction of the leader's motion to plan the follower's trajectory. However, such predictions must be accurate to guarantee string stability, which represents an ongoing challenge for machine learning approaches. This issue can be circumvented by simply incorporating the leader's history, which follows from Newell's car-following (CF) model where a trajectory under congestion corresponds to a temporal–spatial shift of the leader's past trajectories. By leveraging this insight, this paper develops a family of MPC models based on Newell's CF model, labeled Newell MPCs, which are safe and can reduce traffic congestion. Specifically, We first present baseline Newell MPCs to replicate the original Newell's CF model, including the X b o u n d -Model, which uses the shifted leader trajectory as an upper envelope; and the X r e f -Model, which adopts the shifted leader trajectory as a reference to avoid the issue of infeasible solution triggered by hard constraints. To further improve the control performance, we propose the X V -Model which uses the leader speed history as an additional reference to enhance the model robustness and regulate speed over/under-shootings. In addition, we extend the single-leader Newell's model through incorporating multiple leaders and propose the X m u l -Model, which can achieve driver anticipation, and correspondingly reduce reaction time and improve string stability. Finally, based on the X V -Model, we present two additional extensions: (i) the X V r e l a x -Model, which incorporates driver relaxation to achieve smooth response to merging traffic; and (ii) the X V s s -Model, which achieves strict string stability to further dampen traffic oscillations. The proposed Newell MPCs are tested using both numerical simulations and field studies on a stock 2019 Honda Civic using Openpilot and Comma.ai; the source code is available at https://github.com/HaoZhouGT/openpilot. • A family of MPC ACCs based on Newell's CF model. • Leader history can outperform prediction for string stability. • Road test of Newell MPC on a production ACC car model. [ABSTRACT FROM AUTHOR]

• Reinforcement learning (RL)-based longitudinal control for absorbing shockwaves. • Cooperative control benefits both ego and surrounding vehicles. • RL controllers trained on HighD trajectories generalize well to commercial ACC data. • Taking information from two lead vehicles further improve RL model performance. Stop-and-go traffic poses significant challenges to the efficiency and safety of traffic operations. In this study, a cooperative longitudinal control based on Soft Actor Critic (SAC) Reinforcement Learning (RL) is proposed to address this issue. The reward function is carefully designed to consider vehicle cooperation and to achieve three main objectives: safety, efficiency, and oscillation dampening. A global performance metric for oscillation dampening is proposed to evaluate the developed RL and other baseline models. Depending on the number of preceding vehicles that can share maneuver information, two models RL-1 and RL-2 are proposed and compared with human driven (HD) and an adaptive cruise control (ACC) model using the HighD and simulated data. It is found that with information from additional preceding vehicles, RL-2 can dampen shockwaves more efficiently. Specifically, RL-1 and RL-2 decrease traffic oscillation by 15%-36% and 15%-42%, respectively, while HD amplifies the oscillation by 14–37%. The ACC model can also dampen shockwaves but is not as effective as RL-1 and RL-2. The two RL control methods are further evaluated based on data collected using a commercial Model X vehicle. Compared with the commercial Model X ACC vehicle in some controlled settings, the proposed RL methods can better dampen the stop-and-go waves by generating smaller oscillation growth, overshooting, and average acceleration/deceleration rate change, suggesting that they can generalize well in a new but similar environment. Finally, the RL methods are evaluated considering a platoon of vehicles with different RL penetration rates. The results show that they consistently outperform HD and ACC in dampening shockwaves. [ABSTRACT FROM AUTHOR]

Commercial adaptive cruise control (ACC) systems are bi-level: an upper-level planner decides the target trajectory and the low-level system executes it. Existing literature on ACCs mostly focus on the planner algorithms or the actuator delay, while the transition process between them, e.g. the low-level control design and its impact are often ignored. This paper tries to fill this gap by digging into the codebase of a recent open-source self-driving system, Openpilot (OP), Comma.ai, from which we extract and formulate the algorithms at both the upper and lower levels. For linear ACCs, the paper extends the transfer function analysis from planners only to full control loops and investigates the impact of slow/fast low-level control on the overall string stability (SS). For MPC ACCs, it studies their planning characteristics based its optimization objectives and approximates the low-level impact using an ODE approach. We find that low-level control has a significant impact on the overall SS of ACCs: (i) slow low-level control undermines SS under small frequencies and improves SS given large frequencies for linear systems, (ii) MPC features a varying gain throughout an oscillation, where the fast low-level control typically results in a 'fast-slow' changing process of the MPC gain, which benefits the SS, whereas the slow low-level control leads to a 'slow-fast' varying gain which undermines the SS, (iii) slow low-level control are common as they arise from comfort-oriented control gains, from a "weak" actuator performance or both, and (iv) the SS is very sensitive to the integral gain under slow low-level control for both PI and PIF controllers. Overall, the study recommends fast low-level control for ensuring vehicular SS to reduce traffic congestion, considering that large congestion waves usually feature both small frequencies and large amplitudes, although slow controllers could perform even better provided a short and small leader perturbation. The findings of this paper are verified both numerically and experimentally. For the first time in the literature we implement custom ACC algorithms on market cars, and achieve SS on open roads with a random leader by only tuning the low-level controllers. The source code is shared at https://github.com/HaoZhouGT/openpilot to support on-road experiments of arbitrary car-following models, which may be of interest to other studies. • Open-source ACC algorithms from Openpilot, Comma.ai with focus on low-level control. • Slow low-level control undermines string stability even if the planner is stable. • Slow low-level control can arise from control gains, gas/brake systems, or both. • A new experimental method to run custom ACCs on commercial vehicles. [ABSTRACT FROM AUTHOR]

• We develop a unifying framework to unveil car-following (CF) behavior of AVs. • Physical mechanisms under different control paradigms are analyzed and explained. • Traffic wide impacts resulting from CF behavior of multi-class AVs are explored. • Convolved Multivariate Gaussian Process (MGP) is designed to predict the CF behavior. This paper develops a unifying framework to unveil the physical car-following (CF) behaviors of automated vehicles (AVs) under different control paradigms and parameter settings. The proposed framework adopts the flexible asymmetric behavior (AB) model to reveal the control mechanisms and their manifestation in the physical CF behavior, particularly their response to traffic disturbances. A mapping relationship between the AB model parameters and control parameters is then obtained to understand the range of CF behavior possible. Finally, a predictive modeling approach based on a logistic classifier coupled with a convoluted Multivariate Gaussian Process (MGP) is designed to predict the CF behavior of an AV. Analysis of two well-known controllers, linear state-feedback and Model Predictive Control (MPC), show how the proposed framework can uncover the CF mechanisms and provide insights into traffic-level disturbance evolution. The proposed analysis framework remains scalable and can be applied to a variety of controllers. Ultimately, it can guide AV control design that is not myopic, but considers traffic-level performance. [ABSTRACT FROM AUTHOR]