Connected Cruise Controller Systems

• Connected Cruise Controller is a longitudinal control system that augments traditional cruise control with V2V/V2I data to improve safety, efficiency, and traffic flow.
• The systems employ diverse architectures like predecessor-following and bidirectional exchange to enhance string stability and responsiveness over sensor-only controls.
• Experimental implementations have demonstrated significant energy savings and improved stability, though challenges remain with communication delays and mixed traffic integration.

Connected Cruise Controller denotes a class of longitudinal control systems that augment cruise-control or car-following functions with vehicle-to-vehicle or vehicle-to-infrastructure information. Across the literature, the label is used in more than one sense: it may refer broadly to connected longitudinal control in connected and automated vehicles, more narrowly to mixed-traffic controllers for a single connected automated vehicle among human-driven vehicles, or to related formulations such as Leading Cruise Control, in which a connected and autonomous vehicle adapts to preceding traffic while also influencing following traffic (Wang et al., 2018, Wang et al., 2018, Wang et al., 2020).

1. Terminology and scope

The literature distinguishes Connected Cruise Controller from both sensor-only Adaptive Cruise Control and fully cooperative platooning, but the boundary is not uniform across papers. A review of CACC systems presents CACC as an extension of ACC that integrates V2V connectivity with on-board sensing, while mixed-traffic CCC formulations emphasize a single connected automated vehicle using information from connected vehicles ahead, GPS positions, or infrastructure signals without requiring a fully automated platoon (Wang et al., 2018, Alan et al., 2022).

Paradigm	Main information source	Typical emphasis
ACC	On-board sensing of the immediate predecessor	Desired gap and speed regulation
CACC	V2V predecessor/leader state and acceleration	Smaller headways, platooning, string stability
CCC	V2V/V2I augmentation in mixed traffic	Safety, smoothing, efficiency
LCC	“Looking ahead” and “looking behind”	Attenuating downstream perturbations and smoothing upstream flow

ACC is consistently defined as longitudinal control based on on-board sensing of the preceding vehicle only. CACC adds communication of speed, acceleration, position, inter-vehicle distance, or time gap, and is associated with predecessor-following, predecessor–leader following, or multi-predecessor topologies. CCC, by contrast, is used for several mixed-traffic and connectivity-augmented variants: safety-oriented following of a connected vehicle ahead, energy-aware response to multiple vehicles ahead, V2I ecological cruise control with signal timing, and controllers that remain operational under sensing loss or communication uncertainty (Wang et al., 2018, Lin et al., 2019, Bae et al., 2018).

Leading Cruise Control sharpens this distinction. It retains car-following with respect to preceding traffic, but explicitly incorporates the effect of the connected automated vehicle on the vehicles behind. In that formulation, a single CAV embedded in human-driven traffic is treated as a mobile leader rather than only as a follower, and the downstream and upstream roles of the controller are both made explicit (Wang et al., 2020).

2. Modeling foundations and spacing policies

Most connected cruise formulations start from longitudinal point-mass kinematics,

$$\dot{x}_i(t)=v_i(t), \qquad \dot{v}_i(t)=a_i(t), \qquad a_i(t)=u_i(t),$$

combined with a constant time-headway spacing policy,

$$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$

which is the standard analytical backbone for frequency-domain string-stability analysis, consensus protocols, and MPC-based designs (Wang et al., 2018).

Mixed-traffic CCC and LCC models often use linearization around an equilibrium $(s^*,v^*)$. For human-driven vehicles,

$$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$

while the CAV dynamics are

$$\dot{\tilde{s}}_0(t)=\tilde{v}_{-1}(t)-\tilde{v}_0(t),\qquad \dot{\tilde{v}}_0(t)=u(t).$$

This representation underlies controllability, observability, and head-to-tail transfer-function analysis in LCC and related mixed-traffic controllers (Wang et al., 2020).

Higher-fidelity connected cruise models explicitly include actuator dynamics. An energy-efficient EV platooning study uses a third-order switched model,

$$\dot{x}=v,\qquad \dot{v}=a,\qquad \dot{a}=-\gamma(u)a+\beta(u)u,$$

with distinct motoring and regenerative-braking parameters identified from Mustang Mach-E experiments. A connected automated truck model introduces rolling resistance, aerodynamic drag, power limits, and actuation delay through

$$\dot{h}(t)=v_1(t)-v(t),\qquad \dot{v}(t)=-f(v(t))+\tilde{a}(t),\qquad \tilde{a}(t)=\mathrm{sat}(u(t-\sigma)),$$

so that energy-aware CCC tuning can be tied directly to physical power demand (Faghihian et al., 9 Apr 2026, Shen et al., 2022).

Spacing policies also vary. Besides constant time headway, the literature uses piecewise linear range policies such as

$$V(h)=\min\{v_{\max},\max\{0,\kappa(h-d_{\mathrm{st}})\}\},$$

IDM-style desired gaps of the form $H(v)=d+\tau v$, and safety envelopes based on stopping distance, time headway, or time to conflict. This plurality of models reflects different priorities: analytical tractability, actuator realism, traffic-wave attenuation, or energy-aware operation (Shen et al., 2022, Molnar et al., 2023).

3. Connectivity architectures and controller families

The dominant information-flow topologies are predecessor-following, predecessor–leader following, two-predecessor following, multi-predecessor following, and bidirectional exchange. The review literature treats these as core architectural choices because they determine what preview information is available and how disturbance attenuation can be achieved under communication constraints (Wang et al., 2018).

In mixed traffic, CCC often uses upstream information only. A representative safety-oriented truck controller is

$$u_{\mathrm{CCC}}(h,v,v_1)=A(h)\,[V(h)-v]+B(h)\,[W(v_1)-v],$$

where V2V speed and GPS-based positions provide the connectivity-based headway and lead-vehicle speed. In this formulation, the controller reduces to constant-speed cruise when headway exceeds a threshold, but can also engage beyond radar line-of-sight and when the connected vehicle is not the closest vehicle (Alan et al., 2022).

Richer CCC laws use multiple upstream vehicles. Reactive connected cruise control is written as

$$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$0

where $$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$1 are designed response delays to beyond-line-of-sight references. Predictive connected cruise control embeds the same connectivity into an MPC that predicts the motion of the immediate predecessor by propagating the measured motion of a connected vehicle farther ahead through a car-following model across hidden vehicles (Shen et al., 2022).

LCC extends the architecture in the opposite direction by explicitly including following vehicles:

$$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$2

The additional terms associated with $$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$3 make the controller “look behind,” which is the key structural difference from CCC variants that only improve the CAV’s own following behavior (Wang et al., 2020).

Other controller families exploit specific connected data sources. A 5G connected ACC keeps the commercial ACC core intact while modulating the relative-speed input to spacing control using lead-vehicle speed, acceleration, and inverse time-to-collision, and further adapts acceleration bounds and time gap using Pirelli Cyber Tyre friction estimates (Leo et al., 2022). When radar loses the target on curves or hills, another CACC variant replaces direct range sensing with a V2V/localization/map-based approximation of inter-vehicular arc length along the lane centerline rather than reverting to constant-speed cruise (Lin et al., 2019).

4. Stability, controllability, and safety

String stability is the canonical performance notion for connected cruise control. In review form it is stated as

$$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$4

while mixed-traffic LCC studies use the head-to-tail transfer function

$$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$5

The shift from local vehicle stability to string stability is fundamental: internal Lyapunov stability alone does not preclude amplification of disturbances along the vehicle string (Wang et al., 2018, Wang et al., 2020).

LCC contributes a controllability result absent from classical ahead-looking CCC. Under the mild parameter condition

$$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$6

the subsystem composed of the CAV and the following HDVs is controllable, whereas the preceding-vehicle subsystem remains uncontrollable because of uni-directional car-following dynamics. This formalizes the claim that a single connected automated vehicle can actively govern the motion of its followers in mixed traffic (Wang et al., 2020).

Dynamic information-flow topology creates a different stability problem. In CACC-DIFT, link failures switch the controller among CACC1, CACC2, CACC3, and ACC modes. The sufficient string-stability condition for the two-predecessor mode is

$$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$7

whereas ACC fallback requires

$$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$8

This gap explains why retaining a degraded CACC mode under partial communication loss can substantially outperform immediate fallback to sensor-only ACC (Gong et al., 2018).

Safety analysis has increasingly been formalized through control barrier functions. For a follower with spacing $$s_i^*(t)=d_0+h\,v_i(t), \qquad e_i(t)=s_i(t)-s_i^*(t),$$9, ego speed $(s^*,v^*)$0, and leader speed $(s^*,v^*)$1, the main barrier candidates are

$(s^*,v^*)$2

$(s^*,v^*)$3

$(s^*,v^*)$4

These encode minimum distance, time headway, and time-to-conflict, respectively, and enable safe sets of the form $(s^*,v^*)$5 to be rendered forward invariant by a CBF-QP safety filter (Molnar et al., 2023).

Response lag complicates this picture. A first-order lag in the CAV acceleration dynamics reduces both safety and stability margins, and the lag-aware CCC analysis shows that safe operation requires plant and head-to-tail string stability in most cases. The paper derives a critical lag value $(s^*,v^*)$6 above which safe nominal CCC gains do not exist, and then uses an extended CBF construction to synthesize safety-critical CCC that preserves safety even when lag is present (Chen et al., 2024).

5. Data-driven, energy-aware, and resilient designs

A major recent direction is the replacement of explicit traffic models by behavioral or data-driven descriptions. DeeP-LCC constructs a non-parametric predictive controller for mixed traffic using Willems’ fundamental lemma. With measured input/output data and Hankel matrices, future trajectories satisfy

$(s^*,v^*)$7

and the controller solves a regularized QP with spacing and acceleration constraints. In simulations with $(s^*,v^*)$8 followers and $(s^*,v^*)$9 CAVs, DeeP-LCC achieved average real cost $$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$0 versus MPC $$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$1, MSVE reductions up to 43.82%, fuel-consumption reductions up to 24.69% in emergency braking, and mean solve time $$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$2 ms (Wang et al., 2022).

Data-driven CCC has also been used explicitly for energy optimization. A safe and efficient data-driven CCC identifies upstream speed spectra from real traffic, tunes the gains in

$$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$3

and then applies a CBF-based safety filter. The reported outcome is an energy reduction of more than 10% compared to standard non-connected adaptive cruise control, together with a quantified trade-off: the safety filter always removes safety violations, but in some CCC settings it slightly worsens energy because it can overconstrain accelerations and decelerations (Xiao et al., 29 Jul 2025).

Model-based energy-aware CCC remains active as well. For EV platoons, a Lyapunov-based third-order CACC controller achieved up to 38.5% energy savings versus a baseline CACC and maintained string stability at $$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$4 s. For a connected automated truck in mixed traffic with lean penetration of connected vehicles, spectral tuning of a delayed multi-leader CCC produced nonlinear-model energy reductions of 16.44% to 18.11% versus ACC, and the additional delays introduced for distant connected vehicles yielded average energy savings of approximately 1.6–2.0% beyond CCC without delay (Faghihian et al., 9 Apr 2026, Shen et al., 2022).

Connectivity also introduces an attack surface. RACCON addresses V2V attacks on CACC by combining an on-board predictor, comparator, response estimator, and plausibility checker. In the reported experiments, RACCON maintained time headway in the desired range $$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$5 for impactful attack classes where naïve CACC became unsafe, while blanket degradation to ACC remained safe but incurred substantial efficiency loss (Boddupalli et al., 2021).

6. Experimental validation, applications, and limitations

Physical validation spans small robots, production-scale trucks, and vehicle-in-the-loop platforms. A recent tangible platooning platform used four Ubiquity Magni robots with ROS communication, high-level control at approximately 10 Hz, low-level motor control at approximately 50 Hz, and a measured uniform communication delay of $$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$6 s. With $$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$7, $$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$8, $$\dot{\tilde{s}}_i(t)=\tilde{v}_{i-1}(t)-\tilde{v}_i(t),\qquad \dot{\tilde{v}}_i(t)=\alpha_1 \tilde{s}_i(t)-\alpha_2 \tilde{v}_i(t)+\alpha_3 \tilde{v}_{i-1}(t),$$9, $$\dot{\tilde{s}}_0(t)=\tilde{v}_{-1}(t)-\tilde{v}_0(t),\qquad \dot{\tilde{v}}_0(t)=u(t).$$0 s, $$\dot{\tilde{s}}_0(t)=\tilde{v}_{-1}(t)-\tilde{v}_0(t),\qquad \dot{\tilde{v}}_0(t)=u(t).$$1, $$\dot{\tilde{s}}_0(t)=\tilde{v}_{-1}(t)-\tilde{v}_0(t),\qquad \dot{\tilde{v}}_0(t)=u(t).$$2 s, and $$\dot{\tilde{s}}_0(t)=\tilde{v}_{-1}(t)-\tilde{v}_0(t),\qquad \dot{\tilde{v}}_0(t)=u(t).$$3 m, both simulations and field tests satisfied the derived internal and string-stability conditions; the experiments also showed that delays and packet losses increased as platoon size grew beyond about three vehicles (Dutta et al., 30 Jul 2025).

Full-scale mixed-traffic validation has been demonstrated on a connected automated truck. A control-barrier-function safety filter integrated an energy-optimal predictive cruise controller with a safety-oriented CCC, yielding 18% ± 3% to 5% energy savings versus CCC on-track and 4.3% savings versus ACC/CCC on a public highway, while headway remained positive and no safety violations were reported (Alan et al., 2022). In another V2I setting, an ecological ACC implemented on a real PHEV with controller-in-the-loop and environment simulation achieved 41.0% equivalent fuel reduction and 32.91% lower wheel energy than an ACC-only baseline, while avoiding front collisions and traffic-light violations and not stopping at any red light (Bae et al., 2018).

Connected cruise concepts have also been validated in niche but operationally important scenarios. During radar target-detection loss, a map/localization/V2V approximation of lane-center arc length maintained actual time-gap averages of 0.8000 s with standard deviation 0.0264 s in approximation-only operation and 0.8005 s with standard deviation 0.0359 s in switching operation (Lin et al., 2019). At a 2-to-1 lane reduction, a random mixture of 40% connected vehicles improved flow past the bottleneck by 52%, from $$\dot{\tilde{s}}_0(t)=\tilde{v}_{-1}(t)-\tilde{v}_0(t),\qquad \dot{\tilde{v}}_0(t)=u(t).$$4 veh/s to $$\dot{\tilde{s}}_0(t)=\tilde{v}_{-1}(t)-\tilde{v}_0(t),\qquad \dot{\tilde{v}}_0(t)=u(t).$$5 veh/s, by inducing earlier lane changes and delaying self-organized congestion (Davis, 2015). For pairs of connected automated vehicles enclosing packets of human-driven vehicles, connected operation achieved head-to-tail string stability at about 10% penetration, whereas disconnected ACC-only operation required about 30% penetration (Guo et al., 2022).

The main limitations are consistent across the literature. Many analyses are restricted to single-lane longitudinal motion, linearization near equilibrium, ideal or only mildly imperfect communications, homogeneous platoons, or one-CAV-in-HDVs settings. Delay, packet loss, sensor attacks, heterogeneous vehicle dynamics, multi-lane coordination, and formal recursive feasibility for data-driven external-input settings are frequently left to future work. This suggests that connected cruise control is already a mature analytical topic, but still an active systems-integration problem at the intersection of control theory, communication reliability, traffic flow, and safety certification (Wang et al., 2020, Wang et al., 2022, Chen et al., 2024).