Generalized Vicsek Model

• The generalized Vicsek model is a family of nonequilibrium models where self-propelled agents update their headings from local interactions under noise, extending the classic flocking framework.
• It incorporates modifications such as varied noise prescriptions, individualistic updates, and multi-species interactions to trigger continuous, discontinuous, and hybrid order–disorder transitions.
• Extensions involving confinement, topology, and added internal states reveal phenomena like bidirectional order, stripe formation, synchronized rotations, and scale‐free chaos in active matter.

The generalized Vicsek model denotes a family of nonequilibrium models of collective motion built by extending the Standard Vicsek Model (SVM) while retaining its core mechanism: self-propelled agents move at constant speed and update their headings from local information under noise. Across this family, the generalizations modify the heading rule, the statistics of the angular perturbation, the interaction topology, the species structure, the order of interactions, the confinement geometry, or the internal state of each agent. These modifications preserve the status of the Vicsek framework as a minimal theory of flocking, but they also produce qualitatively different macroscopic phenomena, including discontinuous and hybrid order–disorder transitions, coexistence and hysteresis, bidirectional order, striped demixing, scale-free chaos, synchronized rotation, and jamming (Baglietto et al., 2013, Clusella et al., 2021, León et al., 22 Dec 2025, Lardet et al., 22 Mar 2025, González-Albaladejo et al., 2023, Navoret et al., 17 Dec 2025).

1. Standard structure and baseline formulation

In its standard off-lattice form, the Vicsek model consists of $N$ point-like self-propelled particles moving at constant speed $v_0$ in a two-dimensional square domain with periodic boundary conditions. Each particle $i$ carries a position $x_i^t \in \mathbb{R}^2$ and a heading $\theta_i^t$, with velocity

$v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$

Interactions are metric: particle $i$ aligns to neighbors within a disk of radius $R_0$. In the synchronous forward-update scheme, the heading is updated through the argument of the local mean orientation plus angular noise, and the position is then advanced using the updated heading: $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$ The corresponding flocking order parameter is

$$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$

with $v_0$0 in an ordered state and $v_0$1 in a disordered state (Baglietto et al., 2013).

A widely used general formulation keeps the same position update but replaces the angular perturbation by a random variable $v_0$2 drawn from a probability density $v_0$3, where $v_0$4 is the local polarization magnitude and $v_0$5 is a control parameter: $v_0$6 This formulation includes additive angular noise, vectorial noise, and multiplicative noise prescriptions as special cases, and it makes explicit that generalized Vicsek-like models differ primarily through the statistics and state dependence of the heading perturbation (Clusella et al., 2021).

The baseline SVM already exhibits a noise-driven collective transition. For the parameter set $v_0$7, $v_0$8, $v_0$9, $i$0, forward update with periodic boundaries, the transition is continuous and occurs at $i$1 (Baglietto et al., 2013). Generalized Vicsek models depart from this baseline not by abandoning local alignment, but by changing the microscopic decision rule around it.

2. Noise prescriptions and individualistic updates

One major branch of generalization changes the heading update while preserving constant speed, local metric interactions, and the orientational symmetry of the baseline model. The $i$2-extended Vicsek model introduces a probability $i$3 that a particle behaves “individualistically” rather than “gregariously.” With probability $i$4, the heading follows the standard alignment rule with angular noise; with probability $i$5, the particle chooses an independent random direction $i$6. All other parameters, including $i$7, $i$8, the interaction range, and the dimensionality, remain unchanged (Baglietto et al., 2013).

This seemingly minimal change has a strong effect on phase behavior. In the small-noise regime, a relatively small $i$9 of around $x_i^t \in \mathbb{R}^2$0 is sufficient to drive the system from an ordered Vicsek-like phase to a disordered phase, with $x_i^t \in \mathbb{R}^2$1 at $x_i^t \in \mathbb{R}^2$2, $x_i^t \in \mathbb{R}^2$3, $x_i^t \in \mathbb{R}^2$4, $x_i^t \in \mathbb{R}^2$5, and $x_i^t \in \mathbb{R}^2$6. The transition is discontinuous rather than continuous. The evidence consists of a sharp decrease of $x_i^t \in \mathbb{R}^2$7, a pronounced negative minimum of the Binder cumulant

$x_i^t \in \mathbb{R}^2$8

bimodal order-parameter distributions $x_i^t \in \mathbb{R}^2$9, and metastable switching between ordered and disordered states in time series near $\theta_i^t$0 (Baglietto et al., 2013).

A more general route to modification replaces the uniform angular noise by arbitrary symmetric, zero-mean angular PDFs $\theta_i^t$1, possibly depending on the instantaneous local polarization $\theta_i^t$2. In this framework, vectorial Vicsek noise can be re-expressed as a non-standard multiplicative angular PDF depending on $\theta_i^t$3, and two additional multiplicative models are obtained from wrapped Gaussian and bivariate Gaussian distributions. In mean field, the relevant quantity is the mean resultant length

$\theta_i^t$4

For global interactions, vectorial noise and wrapped Gaussian noise generate first-order hybrid transitions, whereas the bivariate Gaussian case yields a second-order transition. For metric short-range interactions, however, all considered multiplicative noise prescriptions display a discontinuous transition with a coexistence region, consistent with the original formulation of the Vicsek model (Clusella et al., 2021).

These results clarify that “noise” in Vicsek-like systems is not a single perturbative ingredient. Its angular law, and especially whether it couples multiplicatively to local order, can determine the order of the transition under mean-field or network coupling, while density–order feedback in metric space can override mean-field predictions and restore discontinuous banding (Clusella et al., 2021). The $\theta_i^t$5-extension sharpens this point further: even preserving $\theta_i^t$6 symmetry, metric range, and dimensionality is not sufficient to preserve the order of the collective transition (Baglietto et al., 2013).

3. Higher-order, multi-species, and antagonistic interactions

A second broad class of generalized Vicsek models alters the interaction structure itself. One example derives from local conformity: the influence weight of neighbor $\theta_i^t$7 on agent $\theta_i^t$8 depends on how well $\theta_i^t$9’s heading matches the local majority around $v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$0. With weights

$v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$1

the heading dynamics decomposes into a pairwise Vicsek term and a genuine three-body term,

$v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$2

with $v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$3 and $v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$4. This higher-order Vicsek model supports a bidirectionally ordered phase in which two groups move in opposite directions, characterized by large nematic order $v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$5 and reduced polar order $v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$6. It displays continuous transitions when $v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$7 and discontinuous, bistable transitions when $v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$8, with explicit mean-field thresholds $v_i^t = v_0 (\cos \theta_i^t, \sin \theta_i^t).$9 for the loss of stability of disorder and $i$0 for abrupt loss of order when $i$1 (León et al., 22 Dec 2025).

Multi-species generalizations replace a single alignment channel by species-dependent intra- and interspecies couplings. In the two-species Vicsek model with “unfriendly” interactions, particles of species $i$2 and $i$3 align with their own species and anti-align with the other. At the microscopic level this is encoded by a spin factor $i$4 in the heading update. The model retains the liquid–gas-like transition and micro-phase separation of the one-species Vicsek model, but the coexistence region contains two distinct dynamical states: PF (parallel flocking), in which the dense bands of the two species propagate in the same direction, and APF (antiparallel flocking), in which they move in opposite directions. In the low-density part of coexistence, PF and APF perform stochastic transitions from one to the other, and the dwell times show a pronounced crossover determined by the ratio of the band width and the longitudinal system size (Chatterjee et al., 2022).

A more general two-species formulation with reciprocal intra- and interspecies couplings $i$5 and $i$6 shows that anti-alignment does not merely suppress order. For $i$7 and $i$8 with $i$9, the model forms flocking stripes: robust, system-spanning traveling stripes of alternating species with strong global polar order. For $R_0$0 and $R_0$1 with $R_0$2, it forms antiparallel flocking stripes with global nematic order. Further regimes include parallel flocking, antiparallel flocking, phase-separated nematic stripes, mixed nematic stripes, independent flocking, independent nematic ordering, and a disordered hyperuniform phase. The stripe wavelength scales linearly with the interaction radius $R_0$3 and is reported numerically as $R_0$4 (Lardet et al., 22 Mar 2025).

These multi-species constructions admit kinetic extensions. A Smoluchowski theory for $R_0$5-species active mixtures with species-dependent alignment matrix $R_0$6 predicts both long-wavelength flocking instabilities and finite-wavelength Turing–Hopf instabilities. In the binary case it reproduces flocking stripes, antiparallel flocking stripes, and nematic stripes, and it predicts a selected length scale $R_0$7 for $R_0$8, matching particle simulations with measured $R_0$9. In cyclic $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$0-species couplings, parity determines the internal stripe organization: odd $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$1 produces distinct chasing stripes for all species, whereas even $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$2 reduces to an effective two-species phenomenology (Lardet et al., 22 Dec 2025).

Taken together, these results shift the generalized Vicsek model from a purely pairwise polar-alignment paradigm to a broader active-matter framework in which triplet couplings, antagonistic interactions, and species networks generate new phases that are inaccessible in the canonical one-species model.

4. Confinement, topology, geometry, and internal-state extensions

Another set of generalizations preserves local alignment but changes the space in which it operates. Harmonically confined Vicsek models replace periodic boundaries by a central restoring term $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$3 inside the alignment field. In two and three dimensions, this modification breaks translational invariance and produces periodic, quasiperiodic, and chaotic attractors instead of the homogeneous ordered and disordered phases of the periodic-box model. On critical curves $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$4, the chaotic state becomes scale free: the correlation length is of the order of the swarm size, the normalized dynamic connected correlation function collapses only for a finite interval of small scaled times, and macroscopic observables obey power laws. In two dimensions, the reported exponents on $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$5 at $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$6 are $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$7, $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$8, $$\theta_i^{t+1} = \mathrm{Arg}\!\left[\sum_{j \in \mathcal{N}_i(t)} e^{i\theta_j^t}\right] + \eta \xi_i^t, \qquad x_i^{t+1} = x_i^t + v_0 (\cos \theta_i^{t+1}, \sin \theta_i^{t+1}) \Delta t.$$9, $$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$0, and $$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$1, while in three dimensions the corresponding values are $$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$2, $$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$3, $$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$4, $$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$5, and $$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$6 (González-Albaladejo et al., 2023, González-Albaladejo et al., 2022).

Interaction topology can also be generalized by imposing limited fields of view and heterogeneous sensing or influence. In a network interpretation, homogeneous rules such as the standard Vicsek model and a restricted-view model produce Erdős–Rényi-like interaction graphs with narrow degree distributions, whereas heterogeneous models with enhanced perception radius (LPVM) or enhanced visibility radius (LVVM) produce Barabási–Albert-like heavy-tailed degree distributions. Across these topologies, the stationary synchronization level $$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$7 is organized by the initial average degree $$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$8 through the stretched-exponential law

$$\phi = \frac{1}{N}\left|\sum_{i=1}^N (\cos \theta_i, \sin \theta_i)\right|,$$9

The law is fitted for SVM, RVM, LPVM, and LVVM over $v_0$00, with $v_0$01 in all cases and often $v_0$02. The paper further reports that LVVM is notably robust, whereas LPVM deteriorates synchronization as heterogeneity increases (Wang et al., 22 Dec 2025).

Generalized Vicsek models may also enlarge the agent state. The Vicsek–Kuramoto model adds an angular velocity $v_0$03 to each agent and couples Vicsek alignment of the direction angle $v_0$04 with Kuramoto-type relaxation of $v_0$05 toward a local mean angular velocity. The microscopic dynamics are

$v_0$06

$v_0$07

Numerical simulations reveal rotating clusters, traveling orientation waves, and globally synchronized rotation, while the hydrodynamic limit supports a gyroscopic term $v_0$08 in the orientation equation (Merino-Aceituno et al., 18 Dec 2025).

At the opposite end of state-space simplification, an Ising-type Vicsek model restricts headings to $v_0$09 on a lattice. Local alignment is encoded through a saturated tanh field $v_0$10, and solitary waves appear as in other Vicsek-type models. In a finite parameter range, the direction of the solitary wave flips abruptly. In one dimension, the average reversal time obeys the empirical law $v_0$11 for $v_0$12, and in two dimensions a bandlike soliton reverses through a turning front that propagates across rows (Sakaguchi et al., 2019).

Finally, geometric and mechanical constraints can be incorporated directly. A Vicsek-like model for collective cell dynamics couples polarity alignment, hard non-overlap constraints implemented by projection onto an admissible velocity cone, soft attraction–repulsion forces, and a relaxation of polarity toward the velocity direction. In periodic domains it recovers an order–disorder transition; in bounded domains it supports coherent rotational flows; at high density or in constrained geometries it exhibits jamming. The model uses the polarity order parameter

$v_0$13

and the normalized mean speed

$v_0$14

showing explicitly that generalized Vicsek dynamics can interpolate between flocking and arrested states when excluded-volume effects are resolved at the microscopic level (Navoret et al., 17 Dec 2025).

5. Kinetic and hydrodynamic theories

The generalized Vicsek model has also become a kinetic and hydrodynamic framework. At the kinetic level, a broad class of Vicsek dynamics on $v_0$15 or $v_0$16 can be written as

$v_0$17

where $v_0$18 is a nonlinear viscosity, $v_0$19 a nonlinear friction, and $v_0$20 a normalized or non-normalized alignment direction built from a general interaction kernel. This formulation includes the original normalized Vicsek choice, for which a singularity may occur when the local flux vanishes. The Cauchy theory establishes local existence and uniqueness under general assumptions on the kernel and coefficients, provides an explicit lower bound on the time of existence $v_0$21, and gives global existence criteria for several important cases, including the non-normalized model and compensated normalized kernels. For a regularized particle system, the mean-field limit is proved, and for short times the probability that the approximated particle dynamics coincide with the original singular dynamics tends to one as $v_0$22 (Briant et al., 2020).

Hydrodynamic limits typically proceed through von Mises or von Mises–Fisher local equilibria and generalized collision invariants. In a two-phase model of moving and non-moving particles, each phase has its own Vicsek alignment operator and the only coupling is switching between phases. The hydrodynamic closure yields coupled PDEs for the densities $v_0$23 and mean directions $v_0$24, together with exchange terms $v_0$25. In the fast-exchange limit, the system reduces to a macroscopic dynamics for the total density and a common direction, with two possible alignment branches: $v_0$26 and $v_0$27 (Navoret, 2014).

Kinetic theory also explains pattern selection in multi-species systems. The Smoluchowski equation for the one-particle distribution $v_0$28 leads, after angular Fourier expansion, to a hierarchy for modes $v_0$29. Retaining the nonlocal metric kernel exactly yields a Bessel factor $v_0$30 in the linearized operator, which is the source of finite-wavelength selection. The resulting instability analysis distinguishes long-wavelength flocking from finite-wavelength Turing–Hopf modes and predicts the stripe wavelength observed in particle simulations (Lardet et al., 22 Dec 2025).

The Vicsek–Kuramoto model supplies an additional hydrodynamic closure in which the equilibrium distribution factorizes into a von Mises law in $v_0$31 and a Gaussian in $v_0$32. The macroscopic equations derived by Generalized Collision Invariants are

$v_0$33

$v_0$34

$v_0$35

The additional conservative law for angular momentum density $v_0$36 and the gyroscopic term $v_0$37 have no analogue in the classical self-organized hydrodynamics of the standard Vicsek model (Merino-Aceituno et al., 18 Dec 2025).

These continuum theories are not merely formal restatements of particle rules. They expose the singular normalized alignment structure, distinguish mean-field from metric-space behavior, identify finite-wavelength instabilities, and clarify how new microscopic ingredients—switching, multi-species coupling, or angular-velocity synchronization—modify the conserved or non-conserved fields of the macroscopic dynamics.

6. Phase-transition phenomenology, diagnostics, and open problems

Generalized Vicsek models are typically classified through a common set of observables: global polarization, species-resolved polarization, nematic order, Binder cumulants, susceptibilities, static and dynamic correlation functions, probability distributions of the order parameter, hysteresis loops, Lyapunov exponents, and, in confined chaotic systems, topological indicators such as Betti numbers (Baglietto et al., 2013, Chatterjee et al., 2022, González-Albaladejo et al., 2023). The continuity or discontinuity of the transition is then inferred from finite-size behavior, coexistence, metastability, hysteresis, or the scaling of the relevant response functions.

Several results recur across otherwise different generalizations. First, preserving orientational symmetry, dimensionality, and interaction range does not fix the transition order: the $v_0$38-extended model preserves $v_0$39 symmetry, metric range, and $v_0$40, yet displays a first-order transition at small noise (Baglietto et al., 2013). Second, multiplicative noise can produce first-order hybrid transitions in mean field, but spatial metric interactions tend to restore discontinuous coexistence even for a mean-field second-order case such as the bivariate Gaussian prescription (Clusella et al., 2021). Third, higher-order or antagonistic couplings generate genuinely new phases—bidirectional order, PF/APF coexistence, flocking stripes, nematic stripes, or cyclic chasing states—rather than merely shifting a critical point (León et al., 22 Dec 2025, Lardet et al., 22 Mar 2025, Lardet et al., 22 Dec 2025). Fourth, confinement and additional internal variables can move the system outside the standard order–disorder taxonomy entirely, toward scale-free chaos or synchronized rotation (González-Albaladejo et al., 2023, Merino-Aceituno et al., 18 Dec 2025).

Open problems are also recurrent. In the $v_0$41 plane of the $v_0$42-extended model, a tricritical point separating the $v_0$43-driven first-order line from the pure-Vicsek continuous line is conjectured but not located (Baglietto et al., 2013). In multiplicative-noise theories, finite-size scaling programs remain incomplete in the spatial metric case (Clusella et al., 2021). In higher-order conformity models, a full hydrodynamic theory including the triplet term is not derived (León et al., 22 Dec 2025). In multi-species systems, the nonlinear saturation and stability of some coexistence phases remain beyond linear theory (Lardet et al., 22 Dec 2025). For network-based formulations, richer interaction physics and time-to-synchronize metrics are natural extensions (Wang et al., 22 Dec 2025). In the rigorous kinetic theory, a full mean-field limit for the original singular normalized particle system remains challenging (Briant et al., 2020). In contact-force models, macroscopic closures that retain jamming and defect-mediated rotation remain an open area (Navoret et al., 17 Dec 2025).

A plausible unifying implication is that the expression “generalized Vicsek model” no longer refers to a single perturbation of the original flocking rule. It refers instead to a broad active-matter class in which local alignment is the common core, but where the macroscopic phenomenology is controlled by how alignment is conditioned—by local order, by independent decisions, by higher-order conformity, by species identity, by confinement, by network topology, by internal angular momentum, or by mechanical contact. In that sense, the generalized Vicsek model is best understood as a research program rather than a single model equation.