Entry–Exit Transition Analysis
- Entry–exit is a transdisciplinary motif that characterizes transition events across system boundaries, detailing timing, costs, and state evolution.
- Applications span surveillance, economic timing, blockchain registries, fluid dynamics, and dynamical systems, providing practical insights into event detection and control strategies.
- Methodologies combine analytical criteria, simulation, and machine learning to quantify transitions, optimize decision-making, and improve system robustness.
Searching arXiv for recent and foundational papers on “entry exit” across the domains represented in the source material. Entry–exit denotes a class of problems concerned with transitions across a boundary, threshold, interface, or regime. In the research literature, the term appears in several technically distinct but structurally related settings: privacy-preserving surveillance of people entering and exiting restricted areas; management of country entry–exit registries; optimal timing of investment, trading, and control decisions; hydrodynamic and hemodynamic phenomena associated with physical entry and exit; and entry–exit relations in slow–fast dynamical systems. Across these settings, the common object is not a single methodology but a recurring analytic motif: the characterization of transition events, their timing, their costs or consequences, and the state evolution before and after the transition (V et al., 2019).
1. Entry–exit as a general analytical motif
In applied computer vision, entry–exit refers to monitoring individuals from outside camera-forbidden areas such as toilets and changing rooms, with the aim of determining entry and exit events, time spent inside, and suspecting transformations in appearance without intruding on privacy (V et al., 2019). In administrative systems, the term denotes recording country border entry and exit records with confidentiality, integrity, and auditability requirements, as in permissioned blockchain registries (Akkad et al., 30 Dec 2025).
In economics and finance, entry–exit frequently denotes the timing of market participation, project activation and liquidation, or firm turnover. Examples include multiple entry problems with forced exits under Poisson shocks (Lempa, 2016), entry and exit with implementation delay under uncertainty (Zhang, 2015), statistical arbitrage with sequential optimal stopping for entry and liquidation (Ning et al., 2023), monopolistic competition with sluggish adjustment of entry and exit (Tanaka, 2019), demand estimation under endogenous product entry and exit (Aguirregabiria et al., 2023), and stationary firm dynamics with entry, exit, and unbounded productivity growth (Stachurski, 2019). A related stochastic-growth formulation studies geometric Brownian motion with intermittent entries and exits, where entry and exit rates jointly shape stationary distributions and first-passage properties (Pal et al., 17 May 2026).
In mechanics and fluid dynamics, entry–exit refers to physical penetration into and emergence from a fluid or to flow through morphological tears. This includes water entry and exit of bodies modeled by potential flow or weakly-compressible SPH methods (Buono et al., 2021, Zhang et al., 2023), and the hemodynamic effects of entry and exit tear size in type B aortic dissection studied with fluid-structure interaction simulation and in vitro 4D-flow MRI (Zimmermann et al., 2023).
In dynamical systems, the entry–exit function, also called a way-in/way-out function, determines delayed loss of stability and exit location near slow manifolds in singularly perturbed systems (Maesschalck et al., 2015, Kaklamanos et al., 2022). This framework has been generalized to multidimensional fast variables and to degenerate turning points (Hsu et al., 2019, Huzak et al., 3 Oct 2025). For jump diffusions, the same vocabulary appears in the classification of entrance and exit at infinity (Doering et al., 2018). This suggests that “entry–exit” is best understood as a transdisciplinary boundary-transition concept rather than a domain-specific term.
2. Computer vision and surveillance uses
The entry–exit surveillance problem was formulated for privacy-preserving monitoring of people entering and exiting camera-forbidden areas from outside the doorway region (V et al., 2019). A central contribution in this line is the EnEx2 dataset, a pseudo-annotated two-camera dataset captured using two IP cameras of resolution at $20$ fps, placed opposite each other $9.5$ m apart and $2$ m above ground, with one camera taking an oblique view of the entrance to avoid capturing the interior (V et al., 2019). The stated motivation was that earlier single-camera datasets exposed only one side of the entrance, complicating person re-identification because of asymmetric or flipped views and occlusions.
The event-detection model in this setting is spatial transition based. Person detection uses a HOG-based deep learning detector, tracking uses Kalman filters, and each track is assigned origin and sink states according to the relation of its initial and final bounding boxes to a predefined entrance region (V et al., 2019). The states are for entry into the scene from outside, for entry into the private area, and for exit from the private area into the scene. The event labels follow ordered pairs: denotes entry, denotes exit, and $20$0 or $20$1 is treated as just appeared (V et al., 2019). Occlusion handling is performed by updating the entrance region when an individual blocks the doorway, and ambiguities are reduced using two-camera views and spatial mapping.
Reported detection results show that EnEx2 achieved recall $20$2, precision $20$3, and F1 $20$4, compared with recall $20$5, precision $20$6, and F1 $20$7 on the earlier EnEx dataset (V et al., 2019). Event detection accuracies on EnEx2 were $20$8 for entry, $20$9 for exit, and $9.5$0 for just appeared, exceeding the corresponding EnEx values $9.5$1, $9.5$2, and $9.5$3 (V et al., 2019). The same model was also evaluated on CAVIAR and PAMELA-UANDES, with lower performance on PAMELA-UANDES attributed to its top-down viewpoint (V et al., 2019).
A related problem is appearance-invariant entry–exit matching, where the goal is to endorse exit of every subject who had entered a private area despite appearance variation (V et al., 2019). The proposed semantic matching model uses visual soft biometric traits—height, body-build, skin complexion, and clothing color—projects each feature via Linear Discriminant Analysis, computes Euclidean distances, and forms an equal-weight collective confidence score
$9.5$4
The method is explicitly positioned as a rank-$9.5$5 narrowing stage before more reliable but costlier biometrics such as gait and face (V et al., 2019). On the EnEx dataset with gallery size $9.5$6, the paper reports rank-1 accuracy $9.5$7 and rank-10 accuracy $9.5$8; with gallery size $9.5$9, rank-10 accuracy is $2$0 (V et al., 2019). A plausible implication is that entry–exit surveillance decomposes naturally into event detection, temporal accountability, and post-transition identity association.
3. Administrative registries and cryptographic record systems
In national border management, entry–exit refers to the creation, storage, and verification of border crossing records. GateChain is a permissioned blockchain application for country entry–exit registry management that records each entry and exit event as a separate block on a distributed ledger (Akkad et al., 30 Dec 2025). The system is designed around immutable record keeping, cryptographically enforced confidentiality and integrity, and real-time, role-based access for authorized institutions such as border police, customs, and immigration (Akkad et al., 30 Dec 2025).
Its consensus protocol is Proof of Authority with a small, fixed validator set identified by public keys, and each block is signed by the responsible authority using ECDSA (Akkad et al., 30 Dec 2025). Personal data in transactions are encrypted using AES-256, represented as
$2$1
while integrity is enforced through hash chaining and block signatures
$2$2
The block structure includes fields such as authority, hash, index, nonce, timestamp, previousHash, signature, transactions, and transactions_root (Akkad et al., 30 Dec 2025).
Performance results reported on $2$3 blocks are approximately $2$4 s per transaction for AES-256 encryption, $2$5 s per transaction for ECDSA signing, and $2$6 s per transaction for ECDSA verification, corresponding to about $2$7 encryptions per second, $2$8 signs per second, and $2$9 verifications per second (Akkad et al., 30 Dec 2025). The paper also reports estimated transaction throughput up to 0 TPS, with signature verification identified as the main bottleneck and described as 1-2 slower than signing (Akkad et al., 30 Dec 2025). The security analysis states that with PoA the system resists up to 3 malicious validators and requires over 4 to break consensus (Akkad et al., 30 Dec 2025).
This administrative use differs from surveillance and dynamical-systems uses in that the primary object is not transition detection from raw observations but secure, auditable registry management after transitions have already been institutionally recognized. The shared concept is nonetheless the same: an entry or exit event is treated as a state-changing record whose semantics depend on chronology, authority, and verifiability.
4. Entry and exit in economics, finance, and optimal timing
A large body of work uses entry–exit to describe optimal timing under uncertainty. In one class of models, an agent may repeatedly invest in a project subject to forced exits at Poisson jump times and permanent loss of re-entry after Bernoulli failure events (Lempa, 2016). The idle-state value is formulated as a multiple stopping problem over entry times 5 and forced exits 6, and a key result is that the optimal entry threshold 7 is independent of the Bernoulli success probability 8 (Lempa, 2016). The threshold is characterized by
9
while catastrophe risk changes the value function through the modified discount rate 0 but does not alter the optimal entry threshold (Lempa, 2016).
A separate real-options formulation analyzes entry and exit decisions with implementation delay under geometric Brownian motion, running cost 1, entry cost 2, exit cost 3, and delay 4 (Zhang, 2015). The paper derives closed-form entry and exit triggers and removes the usual restriction 5. When 6, two entry triggers may emerge,
7
with 8, and it is stated that immediate round-tripping is not always optimal even though an arbitrage opportunity appears to exist (Zhang, 2015). The example given in the paper uses 9, 0, 1, 2, 3, 4, 5, yielding 6, 7, and 8 (Zhang, 2015).
In statistical arbitrage, entry and exit become sequential optimal stopping decisions on mean-reverting spreads. The signature-based framework formulates separate entry and exit stopping problems with transaction costs 9 and discount rates 0, and uses truncated signatures of the augmented path 1 as pathwise features (Ning et al., 2023). The stopping rule is learned through a linear functional on signature space and a loss function over simulated or bootstrapped sample paths. The paper reports that, for the UAL–DAL pair, signature optimal trading achieved daily Sharpe 2 and cumulative return 3, compared with Sharpe 4 and cumulative return 5 for a moving-average baseline (Ning et al., 2023).
Other economic uses shift from individual timing to market structure. In monopolistic competition with sluggish adjustment, the number of firms evolves according to industry total profit, and the paper shows that the steady-state number of firms in the open-loop solution is smaller than in the static equilibrium, whereas the memoryless closed-loop steady state is larger than the open-loop one and may exceed the static equilibrium (Tanaka, 2019). In demand estimation, product entry and exit are endogenous selection variables; a two-step procedure estimates a finite mixture model of entry with latent market types and then controls for propensity scores in BLP or nested-logit demand estimation (Aguirregabiria et al., 2023). In firm dynamics, removing bounded productivity from the Hopenhayn framework yields an exact characterization of stationary equilibrium existence through finiteness of expected lifetime output and a broad result that the firm size distribution has a power law tail (Stachurski, 2019). In an open-system stochastic growth model, intermittent entries at rate 6 and exits at rate 7 produce a stationary population size 8, three moment regimes determined by 9, and an optimal exit rate minimizing mean first-passage time (Pal et al., 17 May 2026).
These formulations share a structural theme: entry and exit are not merely discrete actions but operators that reshape value functions, equilibrium composition, or distributional tails. This suggests that in economics and finance the concept is most naturally understood as a dynamic boundary-control problem.
5. Physical entry and exit: fluids, interfaces, and tears
In computational fluid dynamics, entry and exit refer to motion of bodies through a free surface. One approach develops a fully non-linear potential-flow model for water entry and exit of two-dimensional wedges and axisymmetric cones with prescribed motion, using a boundary element method coupled to a simplified finite element model for thin jets (Buono et al., 2021). The governing problem uses Laplace’s equation for the velocity potential, nonlinear free-surface boundary conditions, gravity, and stabilization procedures such as jet cutting and numerical filtering during exit (Buono et al., 2021). The paper emphasizes that gravity, often neglected in analytical impact models, is rather important, especially in the exit phase, and that the proposed model predicts loads and wetted area better than simplified analytical approaches during exit (Buono et al., 2021).
A second line of work treats water entry and exit with wettability effects using a diffusive wetting model in weakly-compressible SPH (Zhang et al., 2023). The wetting variable satisfies
0
with SPH discretization on outermost solid particles and a resolution scaling 1 (Zhang et al., 2023). The model includes a wetting-coupled identification approach that delays the transition of fluid particles from free-surface to inner states until adjacent solid particles are fully wetted, and applies transport velocity formulation regularization only after that transition (Zhang et al., 2023). Validation includes 3-D sphere entry, 2-D cylinder entry and exit, and the full process from entry to exit, with qualitative and quantitative agreement to experiments and improved reproduction of free-surface “waterfall” breaking during exit (Zhang et al., 2023).
In cardiovascular mechanics, entry and exit appear as tears connecting true and false lumina in type B aortic dissection. A patient-specific study compared a baseline model with both tears about 2, a small-entry-tear model with entry reduced to 3, and a small-exit-tear model with exit reduced to 4 (Zimmermann et al., 2023). Relative to baseline, false lumen flow volume decreased by 5 in FSI and 6 in MRI for a smaller entry tear, and by 7 in FSI and 8 in MRI for a smaller exit tear (Zimmermann et al., 2023). The true-to-false lumen systolic pressure difference changed from 9 mmHg in FSI and 0 mmHg in catheter measurements at baseline to 1 and 2 mmHg with a smaller entry tear, and to 3 and 4 mmHg with a smaller exit tear (Zimmermann et al., 2023). The paper interprets these changes as particularly notable for false-lumen pressurization and supports FSI deployment in clinical studies (Zimmermann et al., 2023).
Here, entry and exit are literal interface-crossing processes. Yet the mathematical pattern remains familiar: state transition is governed by geometry, local flow, and accumulated effects before and after the transition.
6. Entry–exit functions in dynamical systems and stochastic processes
The classical entry–exit function in slow–fast systems describes delayed loss of stability near a critical manifold. For
5
with 6 for 7 and 8 for 9, solutions are attracted to the $20$00-axis for $20$01, remain near it after crossing $20$02, and leave when accumulated repulsion balances prior attraction (Maesschalck et al., 2015). The limiting exit point $20$03 satisfies
$20$04
The same work shows that the return map can be written as a $20$05 function of $20$06, and under a flatness property it becomes $20$07 in $20$08 (Maesschalck et al., 2015).
This theory has been extended to systems with two fast variables and one slow variable when the fast eigenvalues intersect before contraction and expansion balance along any individual eigendirection (Kaklamanos et al., 2022). In that setting the classical formula is insufficient because eigendirections coalesce and exchange stability. The paper derives alternative formulas depending on a parameter $20$09: for $20$10,
$20$11
while for $20$12 and an invariant branch one first determines $20$13 by
$20$14
and then solves
$20$15
for the exit point (Kaklamanos et al., 2022).
A multidimensional generalization treats systems
$20$16
with invariant sets $20$17 linked by a singular closed orbit (Hsu et al., 2019). The entry–exit relation is reformulated on an augmented $20$18 space using integrated instability variables, and periodic orbits arise as fixed points of a composed return map $20$19 (Hsu et al., 2019). The paper uses this framework to establish existence and stability of relaxation oscillations in predator–prey systems with rapid ecological evolutionary dynamics (Hsu et al., 2019).
More recent work addresses degenerate turning points. For planar slow–fast systems with invariant line $20$20 and slow flow $20$21 at $20$22, the case $20$23 admits a well-defined entry–exit relation and a Dulac map smooth in $20$24, whereas for $20$25 additional control parameters are required (Huzak et al., 3 Oct 2025). In the $20$26 case the limiting relation is expressed באמצעות Cauchy principal values: $20$27 (Huzak et al., 3 Oct 2025). This suggests that the entry–exit idea persists even under degeneracy, but its representation becomes increasingly dependent on blow-up coordinates and auxiliary parameters.
A related stochastic-process usage concerns entrance and exit at infinity for stable jump diffusions
$20$28
For $20$29, finite-time explosion is possible; for $20$30, entrance from infinity is possible under explicit integral tests involving $20$31, and for $20$32 a logarithmic criterion appears (Doering et al., 2018). Although conceptually distinct from slow–fast delay, this work again treats entry and exit as boundary classifications governed by integral balance conditions.
7. Unifying themes and domain-specific distinctions
A first unifying theme is boundary localization. In surveillance, the relevant boundary is an entrance region $20$33 in image coordinates (V et al., 2019). In optimal stopping and investment, it is a threshold $20$34 or trigger price $20$35 (Lempa, 2016, Zhang, 2015). In water entry and exit, it is the fluid interface and wetted area (Buono et al., 2021, Zhang et al., 2023). In aortic dissection, it is the geometry of entry and exit tears (Zimmermann et al., 2023). In dynamical systems, it is a neighborhood of a critical manifold or the point at infinity (Maesschalck et al., 2015, Doering et al., 2018).
A second theme is transition timing under delayed or accumulated effects. The delay may be literal, as in implementation lag for project decisions (Zhang, 2015), or implicit, as in delayed loss of stability in slow–fast systems (Maesschalck et al., 2015, Kaklamanos et al., 2022). It may appear as survival and first-passage structure under intermittent exits and re-entries (Pal et al., 17 May 2026), or as time spent inside a private area in entry–exit surveillance (V et al., 2019).
A third theme is the role of side information and system architecture. Two-camera geometry improves observability in EnEx2 (V et al., 2019). Visual soft biometrics narrow candidate galleries for appearance-invariant matching (V et al., 2019). Permissioned blockchain architecture and PoA consensus define admissible entry–exit record flows in GateChain (Akkad et al., 30 Dec 2025). In demand estimation, latent market types and propensity scores encode unobserved selection into product entry and exit (Aguirregabiria et al., 2023).
A fourth theme is the prevalence of integral criteria. Examples include the threshold equation in multiple-entry investment (Lempa, 2016), integral balances in entry–exit functions for slow–fast systems (Maesschalck et al., 2015, Kaklamanos et al., 2022, Huzak et al., 3 Oct 2025), perpetual-integral tests for explosion or entrance at infinity (Doering et al., 2018), and the renewal or first-passage formulas in intermittent-entry GBM (Pal et al., 17 May 2026). This suggests that many entry–exit problems reduce, after suitable transformation, to accumulated balance laws between opposing tendencies such as attraction and repulsion, inflow and outflow, or gain and cost.
The principal distinction across fields is semantic rather than structural. In surveillance and registries, entry and exit are events to be detected, recorded, and verified. In economics and finance, they are decisions or equilibrium mechanisms. In fluids and hemodynamics, they are physical processes shaped by geometry and interface conditions. In dynamical systems, they are asymptotic relations governing delayed departure from unstable sets. The term therefore functions as a shared vocabulary for transition phenomena whose specific mathematics depends on the state space, observables, and admissible controls of the underlying discipline.