Pith. sign in

REVIEW 3 major objections 5 minor 22 references

HenTwin: A Multimodal Digital Twin Framework for Longitudinal Biological State Monitoring in Laying Hens

T0 review · 3 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read HenTwin turns four barn signals into a single flock-state model and claims that a sustained +2 THI heat rise would lift acoustic entropy by 0.54 nats, about one-quarter of its 25-week developmental decline.

desk verdict The paper's headline perturbation number is contradicted by its own transition equation; the dataset and two-tier transfer idea are solid, but the central claim needs a fix. read the letter →

arxiv 2607.28652 v1 pith:WTZHHJF4 submitted 2026-06-19 q-bio.OT cs.AIstat.AP

Yashan Dhaliwal , Shreya Rao , Suresh Neethirajan This is my paper · ORCID
classification q-bio.OTcs.AIstat.AP
keywords digitaltwinprecisionlivestockfarmingstate-spacemodelmultimodalsensingpoultrymonitoringtemperature-humidityindexacousticenergyentropychange-pointdetection
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

HenTwin aims to give precision livestock farming a formal state representation instead of a collection of sensor dashboards. The paper constructs a four-dimensional biological state vector—body surface temperature, acoustic energy entropy, band energy ratio, and optical-flow motion—and fits a stable diagonal transition model S(t+1)=A·S(t)+c+B·u(t) over 25 weeks of multimodal data from 150 laying hens, with the temperature-humidity index treated as an exogenous input so the model can answer intervention questions. Its headline demonstration is a perturbation analysis claiming a sustained +2.0 THI increase raises acoustic entropy by 0.54 nats at steady state, about one-quarter of the observed developmental decline, and a coordinated change-point at weeks 12–14 detected across three independent modalities. If the framework is right, it would give barn managers a quantitative what-if tool, show that transition structure transfers across rooms while environmental sensitivity needs local calibration, and take the first step toward state-aware digital twin inference in poultry.

What carries the argument

The central object is the diagonal state-space transition model S(t+1)=A·S(t)+c+B·u(t)+w(t), with per-modality persistence coefficients on the diagonal of A (BST 0.668, EE 0.831, BER 0.543, motion 0.383) and a contemporaneous input-coupling vector B estimated by regressing each channel on THI deviation. The machinery that carries the argument is the combination of asymptotic stability (spectral radius 0.831 < 1), the analytic steady state S* = c/(1−A) used for initialization, and forward simulation under perturbed input u(t)+δ to compute long-run responses. This is what turns raw sensor features into a twin that supports counterfactual what-if scenarios, change-point monitoring, and a two-ti

What would settle it

Simulate the paper's transition equation with the reported parameters (A_EE=0.831, B_EE=0.271, c_EE=1.277) and a constant +2.0 THI deviation, starting from the steady state, and compare the converged ΔEE to 0.54 nats. The long-run multiplier for the EE channel is 1/(1−0.831)≈5.9, so the equilibrium response should be about 3.2 nats; a converged value of 0.54 nats would confirm the reported calculation, whereas a value near 3.2 nats would show the reported steady-state elevation is not what the model actually produces.

Watch

Extended reading notes

Core claim

The paper's central claim is that a discrete-time linear state-space model can serve as a digital twin of a laying-hen flock's multimodal biological state. The state vector S(t) = [BST, EE, BER, Motion]ᵀ evolves as S(t+1)=A·S(t)+c+B·u(t)+w(t), with A a diagonal persistence matrix whose eigenvalues (0.668, 0.831, 0.543, 0.383) are all inside the unit circle, giving a stable system with a unique steady state. From 25 weeks of data the authors estimate B (EE: +0.271 nats per THI unit, BST +0.300 °C, BER +3.334%, motion −0.037) and use forward simulation to show that a sustained +2.0 THI shock produces a stable long-run entropy elevation of 0.54 nats, roughly one-quarter of the 1.87-nat developm

Load-bearing premise

The load-bearing premise is that the measured link between each biological channel and the temperature-humidity index is a true environmental push that drives future state, not a shared trend or room effect; if that link is contaminated, the simulated responses lose their meaning.

Editorial extensions

If this is right

  • A new barn or room would need only a short local calibration period—re-estimating B and the intercept—rather than a full 25-week longitudinal study, because the transition structure transfers across rooms.
  • Acoustic energy entropy is the slowest-adapting channel (half-life ≈ 3.7 weeks), so thermal effects on vocal complexity persist for weeks; monitoring and alerting should operate on that timescale.
  • The coordinated change point at weeks 12–14 provides a data-driven marker of a system-level developmental transition that managers could use to time management changes.
  • Operator questions such as 'what happens to flock acoustic state if THI stays above 72 for three weeks?' become quantitatively answerable through the model's forward simulation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper's long-run entropy response of 0.54 nats equals B_EE·δ, but for its own transition equation the steady-state multiplier is 1/(1−0.831)≈5.9, so the converged response to a +2 THI shock should be roughly 3.2 nats; checking the simulation code would show whether the reported figure is the contemporaneous or the equilibrated impact.
  • Because THI and the biological channels both decline monotonically across weeks 5–25, the contemporaneous B coefficients may partly capture co-maturation rather than causal environmental forcing; detrending the series before estimation would test whether the couplings survive.
  • The order-of-magnitude difference in acoustic coupling between the two rooms (B_EE 0.271 vs 0.027) despite nearly identical THI inputs suggests B absorbs room-specific acoustics such as microphone placement; a follow-up that swaps microphone locations across rooms would separate room effects from flock response.
  • The weeks 12–14 state transition could be cross-checked against independent indicators like feed intake, egg production onset, or behavioural observations to see whether the detected shift marks a biological milestone or an artifact of the aggregation window.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. HenTwin defines a four-dimensional biological state vector for laying hens (BST, EE, BER, motion) with THI as an exogenous input, estimates a diagonal discrete-time state-space model S(t+1)=A·S(t)+c+B·u(t)+w(t) from 25 weeks of multimodal data (20 weekly transitions in Room 1), and uses this model for steady-state analysis, perturbation simulation, change-point detection, cross-room validation, and leave-one-out cross-validation. The headline quantitative claim is that a sustained +2.0 THI increase produces a stable long-run acoustic entropy elevation of 0.54 nats, roughly one-quarter of the observed 1.87-nat developmental decline. The paper also reports coordinated Pettitt change-points at weeks 12–14, partial cross-room transferability, and a two-tier calibration architecture.

Significance. If the central claims held, this would be a useful formalization of multimodal state dynamics in precision livestock farming, with a clearly specified state vector, an estimated transition model, stability verification, and reproducible notebooks and data (Section VI.F). The paper is commendably explicit about many limitations: 20 training observations, the diagonal restriction on A, farm-wide thermal imaging, and the absence of an observation equation. However, the central perturbation claim is internally inconsistent with the model's own transition equation: the steady-state response to a sustained input shift δ is (I−A)^{-1}Bδ, not Bδ, giving approximately 3.2 nats for EE rather than 0.54 nats. Because this number is the paper's principal claimed contribution—intervention-aware digital twin inference—the result as stated cannot stand.

major comments (3)
  1. [§V.E, Eqs. (9)–(12), Fig. 12] For the model in Eq. (6), a sustained input shift δ changes the steady state by (I−A)^{-1}Bδ, not Bδ. With A_EE=0.831 and B_EE=0.271 (Sections VII.B, VII.C), the long-run EE response to δ=2 is 0.271×2/(1−0.831)=3.21 nats. The paper's assertion in §V.E that 'for this linear model the long-run impact equals B·δ' and Fig. 12b's convergence to 0.54 nats are therefore inconsistent with the model's own transition equation; 0.54 is the one-step impact Bδ. This is load-bearing because the abstract, §VII.D, and §V.F all present 0.54 nats as the stable intervention response.
  2. [§V.D, Eq. (8), §VII.C/G] The B vector is estimated by contemporaneous OLS x_i(t)=B_i·u(t)+k_i+ε(t) and is then used as the forcing coefficient in the transition model for perturbation analysis. The paper acknowledges that when B is estimated jointly with A from the transition equation, intervals are wide and include zero for EE. This is not a minor estimation detail: the contemporaneous slope identifies a static association, not a dynamic forcing parameter. Since THI and EE both decline monotonically over weeks 5–25, the R²=0.725 association may reflect a shared developmental trend. The room comparison reinforces this: nearly identical THI in Rooms 1 and 5 yields B_EE=0.271 vs 0.027 (Fig. 17), attributed to room acoustics, implying B measures room-specific covariation rather than a transferable exogenous coupling. Without a defensible identification of B, the perturbation and intervention claims do not follow.
  3. [§VII.H, Tables VI–VII] Cross-room transfer claims are not supported for the headline channel. Direct Room 1-to-5 transfer of EE yields R²=−467.98 (Table VI); after recalibrating B and intercept, one-step EE R² is −0.14 (Table VII), still below a mean predictor. BST cross-room validation is invalid because BST is a farm-wide average with no room tagging (Section III.C.2). Only BER transfers (R²=0.632). This does not establish 'structural transition parameters are partially transferable' as a general result; it shows one channel transfers and others do not. The two-tier deployment conclusion rests on a very narrow empirical base.
minor comments (5)
  1. [§VII.C] The sentence about the BCa bootstrap excluding zero for all four channels appears twice; remove the duplication.
  2. [§VII.F / Fig. 15] The text reports that the first two principal components explain 83.8% of variance, while the Fig. 15 caption says 83.9%. Please reconcile.
  3. [§V.C, Table II] BER observed mean (7.43%) differs from the analytical steady state (4.44%) by 3 percentage points; calling this 'closely matched' overstates agreement. Please quantify the discrepancy or explain it via the channel's non-stationarity.
  4. [§VII.E, Fig. 14] The BER Pettitt test reports p<0.001 but no test statistic K; include K and the exact p-value for consistency with EE and motion baseline.
  5. [Throughout] There are numerous typographical spacing issues (e.g., 'Fig. 1 5', inconsistent spaces around R² values). A careful copyedit is needed.

Circularity Check

1 steps flagged · score 6.0 of 10

The headline 0.54-nat EE perturbation response reduces to the contemporaneous fitted slope B_EE times a hand-chosen δ, and the paper's own Eq. (6) implies a different long-run value, making the claimed stable long-run prediction a restatement of the fit rather than a model-derived result.

  1. fitted input called prediction [§V.D (Eq. 8), §V.E (Eqs. 9–12), §VII.D (Fig. 12)]
    "For this linear model the long-run impact equals the product of the equilibrium input coupling and the applied shift, B·δ, computed elementwise. ... Under this perturbation, EE converged to a stable long-run increase of 0.54 nats above the unperturbed trajectory."

    B_EE is estimated by OLS regression of EE(t) on the contemporaneous THI deviation over the same 20-week training window (Eq. 8), and the reported long-run EE response is exactly B_EE·δ = 0.271×2.0 = 0.54 nats. The paper asserts rather than derives that the long-run impact equals B·δ; from its own transition equation (6), the equilibrium impact of a sustained δ is (I−A)^{-1}Bδ = 0.271×2.0/(1−0.831) ≈ 3.21 nats. The headline 'prediction' thus reduces to the fitted regression slope multiplied by a hand-selected shock size, and the stated model's dynamics contradict the claimed long-run value.

full rationale

The main circularity is concentrated in the perturbation claim. Section V.D estimates B_i by regressing each state variable on the contemporaneous THI deviation over the same 20-week training window (Eq. 8), and Section V.E then asserts that the long-run impact of a sustained shock is B·δ. Section VII.D reports EE = 0.54 nats, which is exactly 0.271×2.0. So the headline 'prediction' is a re-statement of the fitted slope times a hand-chosen shock, not an independent model output. Worse, the paper's own transition equation S(t+1)=A S(t)+c+B u(t) implies a steady-state response (I−A)^{-1}Bδ = 3.21 nats for EE (A=0.831, B=0.271, δ=2.0), so the reported 'converged long-run' value is inconsistent with the model's own dynamics; the assertion that the long-run impact equals B·δ is definitional rather than derived. I did not find load-bearing self-citation circularity: the cited pilot study and reviews are contextual, and no uniqueness theorem is imported to force the modelling choice.

Assumptions & free parameters 11 free parameters · 7 assumptions · 2 invented entities

The quantitative core rests on four fitted persistence coefficients, four fitted intercepts, four contemporaneous THI slopes, and a hand-picked shock size δ=+2.0. Because the headline perturbation response is B·δ — a fitted slope times a chosen constant — the central 'prediction' is a relabeled regression output. Elements with the most independent grounding are the Pettitt change-points (weeks 12–14), the stability check (diagonal entries <1 by construction), and the cross-room transfer test (which fails badly for EE before calibration, R²=−467.98).

free parameters (11)
  • A_BST persistence = 0.668
    OLS estimate of the first-order autocorrelation of BST from 20 week-pairs (weeks 5–25); enters the long-run multiplier 1/(1−A).
  • A_EE persistence = 0.831
    OLS estimate for energy entropy; the largest eigenvalue (spectral radius 0.831) and the key factor in the discrepancy between B·δ and (I−A)^{-1}Bδ.
  • A_BER persistence = 0.543
    OLS estimate for band energy ratio.
  • A_motion persistence = 0.383
    OLS estimate for motion baseline; lowest persistence channel.
  • Intercept vector c = 8.280 (BST), 1.277 (EE), 2.032 (BER), 0.359 (motion)
    OLS intercepts in the diagonal transition model; together with A they set the steady states S*=c/(1−A).
  • B_BST coupling = +0.300 °C per THI unit (R²=0.462)
    Contemporaneous OLS slope of BST on THI deviation (Eq 8); used directly as the perturbation multiplier.
  • B_EE coupling = +0.271 nats per THI unit (R²=0.725)
    Contemporaneous OLS slope of EE on THI; the headline 0.54-nat result is 2×0.271, i.e., the fitted slope times the chosen shock.
  • B_BER coupling = +3.334 percentage points per THI unit (R²=0.311)
    Contemporaneous OLS slope; drives the 6.67-pp BER response reported 'for completeness'.
  • B_motion coupling = −0.037 units per THI unit (R²=0.327)
    Contemporaneous OLS slope; sign used to support behavioural thermoregulation interpretation.
  • Perturbation size δ = +2.0 THI units
    Hand-selected as the upper bound of the observed weekly THI deviation range (−2.26 to +2.26); the reported response scales linearly with this chosen input, so the headline magnitude is partly chosen.
  • Thermal image curation = 25 curated images per week (of 140–160 acquired)
    Images selected by sharpness, visibility of head/foot, no motion blur; not room-tagged, so one farm-wide BST series is used for both rooms — a curation and aggregation choice affecting the BST channel.
assumptions (7)
  • domain assumption The true dynamics are first-order linear, time-invariant, diagonal, with additive zero-mean Gaussian noise (Eq 6, §V.B).
    Imposes AR(1)-plus-input structure on each channel and asserts cross-variable interactions can be absorbed into residuals; a full 4×4 coupling is rejected for sample-size reasons, not tested.
  • domain assumption THI is strictly exogenous with the stated causal direction (Eq 5, §V.A).
    Justifies treating THI as an un-driven input; if room-level biology feeds back on room microclimate at weekly scales, the perturbation interpretation is invalid.
  • ad hoc to paper The contemporaneous OLS slope of x_i(t) on u(t) identifies the input coupling B_i (Eq 8, §V.D).
    Load-bearing: if the EE–THI slope (R²=0.725) reflects shared monotone developmental trends or room-specific acoustic geometry rather than environmental forcing, the perturbation analysis collapses. The 10:1 Room-1/Room-5 difference in B_EE under near-identical THI inputs is a direct warning sign.
  • ad hoc to paper Retaining only p<0.05 couplings yields the correct B vector, with non-significant entries set to zero.
    Post-hoc variable selection on the same data used for the headline projection (SC excluded at p=0.225; motion reactivity at p=0.172) shapes the state vector and the reported perturbation magnitudes.
  • standard math Pettitt test asymptotic p-values are valid for n≈25 weekly observations and detect a single true shift.
    The non-parametric change-point test (§VI.B) is applied to short, trending series; the week-12–14 'coordinated transition' claim rests on the test's behavior on these series.
  • domain assumption Weekly scalar aggregation of each modality preserves the dynamics relevant to flock state.
    Each raw stream (e.g., 10–12 one-minute audio clips per week; three 15-minute video segments per week) is reduced to one number per week; sub-week dynamics and within-week variance are discarded.
  • domain assumption The same farm-wide BST series represents both Room 1 and Room 5.
    Thermal images were not room-tagged (§III.C.2); the paper acknowledges this and uses one BST series for both rooms, making BST-based cross-room validation circular by construction.
invented entities (2)
  • Four-dimensional multimodal biological state vector S(t)
    purpose: Formal representation of flock biological state across thermal, acoustic, and motion channels, with THI as exogenous input.
    A measurement-derived construct. Its status as a 'biological state' with first-order Markov evolution is asserted, not independently validated — the paper itself defers linkage to validated welfare outcomes to future work (§VIII.D).
  • Digital-twin mirror / twin abstraction layer
    purpose: Virtual flock representation supporting forward perturbation simulation and change-point monitoring, with operator-facing decision support.
    The live, continuously updating mirror and decision-support tier are drawn as dashed components in Fig. 1 and explicitly described as the proposed deployment pathway rather than something implemented and validated here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HenTwin: A Multimodal Digital Twin Framework for Longitudinal Biological State Monitoring in Laying Hens." pith.science (2026). https://pith.science/paper/WTZHHJF4

@misc{pith2026260728652,
  author       = {Pith},
  title        = {Pith review of: HenTwin: A Multimodal Digital Twin Framework for Longitudinal Biological State Monitoring in Laying Hens},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WTZHHJF4}},
  note         = {Machine review of arXiv:2607.28652}
}
read the original abstract

Early-life monitoring in laying hens remains constrained by fragmented single-modality sensing and the absence of formal system-level state representations. HenTwin, a multimodal digital twin framework implemented as a five-layer IoT architecture, formalizes flock-level multimodal biological state dynamics from hatch through 25 weeks of age. A four-dimensional biological state vector integrating body surface temperature, acoustic energy entropy, band energy ratio, and optical-flow-based motion is defined, with the temperature-humidity index treated as an exogenous environmental input to preserve intervention capability. A discrete-time state transition model is estimated from 25 weeks of longitudinal multimodal data collected from 150 Lohmann LSL-Lite hens across five controlled rooms at the Atlantic Poultry Research Centre, Dalhousie University. The estimated transition matrix exhibits modality-specific persistence while remaining asymptotically stable. Perturbation analysis demonstrates that a sustained +2.0 THI increase produces a stable long-run acoustic entropy elevation of 0.54 nats, approximately one-quarter of the entire 1.87-nat developmental decline observed across the study period. Pettitt change-point detection identifies coordinated multimodal developmental state transitions at Weeks 12-14. Cross-room validation suggests that structural transition parameters are partially transferable across rooms, whereas environmental input sensitivity requires room-specific calibration, supporting a two-tier IoT deployment architecture. Leave-one-out cross-validation demonstrates consistent out-of-sample model performance. HenTwin takes a first step toward formal, state-aware digital twin inference in precision livestock farming.

Figures

Figures reproduced from arXiv: 2607.28652 by the authors.

Figure 1
Figure 1. HenTwin five-layer IoT architecture. Multimodal barn signals are acquired by the sensing layer (Layer 1) across four modalities – microclimate (DHT22), thermal (FLIR Cx-5), acoustic (Zoom/Wildlife recorder), and visual (GoPro Hero 13). The edge processing layer (Layer 2) reduces each modality to a compact feature through a dedicated pipeline: the temperature–humidity index (THI) from microclimate, body surface tempe… view at source ↗
Figure 3
Figure 3. Thermal imaging workflow for Body Surface Temperature [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 4
Figure 4. HenTwin acoustic recording array deployed across the study [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figures from the paper (10 more)
Figure 5
Figure 5. Figure 5: Video acquisition setup for behavioural monitoring in Room [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 6
Figure 6. Figure 6: Weekly temperature–humidity index (THI) for Rooms 1 and 5 across the 25-week rearing period, shown against established laying-hen thermal comfort categories (comfort < 70, alert 70–75, danger 76–81, emergency > 81) [22]. Both rooms track closely and decline steadily fr…
Figure 8
Figure 8. Figure 8: Cross-room comparison of Spearman rank correlations for the principal couplings (THI–EE, THI–BER, THI–BST, and BST–EE). The THI–BST and THI–BER associations are of comparable strength in both rooms, whereas the THI–EE coupling is substantially stronger in Room 1 (ρ = 0…
Figure 9
Figure 9. Figure 9: Estimated state transition matrix A of the HenTwin diagonal state [PITH_FULL_IMAGE:figures/full_fig_p013_9.png]
Figure 11
Figure 11. Figure 11: Relationship between the temperature–humidity index and energy entropy across the full 25-week study (weeks 1-25), for (a) Room 1 and (b) Room 5. Each point represents one study week; dashed lines show the ordinary least squares fit, and panels share a common energy-e…
Figure 13
Figure 13. Figure 13: Multimodal state response to a sustained +2.0 THI perturbation, computed with the diagonal state-space model. Each panel shows the observed weekly value (black), the model baseline trajectory under the observed THI series (blue, dashed), and the perturbed trajectory u…
Figure 14
Figure 14. Figure 14: Pettitt change-point detection in the Room 1 state variables: (a) energy entropy, (b) band energy ratio, and (c) motion baseline. A coordinated developmental transition is detected within a narrow window – energy entropy shifts at week 14 (K = 144, p < 0.001; mean 8.2…
Figure 16
Figure 16. Figure 16: Direct cross-room transfer of the Room 1 state-space model applied, without recalibration, to Room 5 observations: (a) body surface temperature, (b) energy entropy, and (c) band energy ratio. In each panel the Room 5 observed series (black) is compared with the Room 1…
Figure 17
Figure 17. Figure 17: Between-room acoustic sensitivity to thermal load. (a) Energy entropy versus the temperature–humidity index for Room 1 (blue) and Room 5 (red), with ordinary least squares fits; the THI–EE coupling is approximately an order of magnitude stronger in Room 1 (B = +0.271 …
Figure 18
Figure 18. Figure 18: Two-tier calibration transfer test, Room 1 model applied to Room 5 (all 25 weeks; THI deviation referenced to the Room 1 training mean). (a) Input coupling coefficients (B) for each shared state variable under three conditions – the trained Room 1 vector, the Room 5 r…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 7 canonical work pages

  1. [1]

    Early life programming in poultry: Recent insights and interventional approaches,

    A.-M. E. Abdel -Moneim, A. M. Shehata, and V. K. Paswan, "Early life programming in poultry: Recent insights and interventional approaches," Frontiers in Veterinary Science , vol. 9, 2023, doi: 10.3389/fvets.2022.1105653

    arXiv 2023

  2. [2]

    Influence of age, time of day, and environmental changes on vocalization patterns in broiler chickens,

    P. Soster, T. Grzywalski, K. Buyse, P. Thomas, and G. Antonissen, "Influence of age, time of day, and environmental changes on vocalization patterns in broiler chickens," Poultry Science, vol. 104, no. 8, p. 105298, 2025, doi: 10.1016/j.psj.2025.105298

    arXiv 2025

  3. [3]

    Reliability, practicability and farmers' acceptance of an animal welfare assessment protocol for broiler chickens and turkeys,

    S. Michaelis, D. Gieseke, and U. Knierim, "Reliability, practicability and farmers' acceptance of an animal welfare assessment protocol for broiler chickens and turkeys," Poultry Science, vol. 103, no. 10, p. 103900, 2024, doi: 10.1016/j.psj.2024.103900

    arXiv 2024

  4. [4]

    Automated techniques for monitoring the behaviour and welfare of broilers and laying hens: Towards the goal of precision livestock farming,

    N. Li, Z. Ren, D. Li, and L. Zeng, "Automated techniques for monitoring the behaviour and welfare of broilers and laying hens: Towards the goal of precision livestock farming," Animal, vol. 14, no. 3, pp. 617 –625, 2020, doi: 10.1017/S1751731119002155

    work page doi:10.1017/s1751731119002155 2020

  5. [5]

    Decoding poultry welfare from sound: A machine learning framework for non -invasive acoustic monitoring,

    V. Manikandan and S. Neethirajan, "Decoding poultry welfare from sound: A machine learning framework for non -invasive acoustic monitoring," Sensors, vol. 25, no. 9, p. 2912, 2025, doi: 10.3390/s25092912. 24

    doi:10.3390/s25092912 2025

  6. [6]

    Digital twins in poultry farming: Deconstructing the evidence gap between promise and performance,

    S. R. Neethirajan, "Digital twins in poultry farming: Deconstructing the evidence gap between promise and performance," Applied Sciences, vol. 16, no. 3, p. 1317, 2026, doi: 10.3390/app16031317

    work page doi:10.3390/app16031317 2026

  7. [7]

    Digital twins for cows and chickens: From hype cycles to hard evidence in precision livestock farming,

    S. Neethirajan, "Digital twins for cows and chickens: From hype cycles to hard evidence in precision livestock farming," Agriculture, vol. 16, no. 2, p. 166, 2026, doi: 10.3390/agriculture16020166

    work page doi:10.3390/agriculture16020166 2026

  8. [8]

    Designing an ISO 23247-compliant hybrid digital twin architecture for industry,

    A. Garro and A. Sorrenti, "Designing an ISO 23247-compliant hybrid digital twin architecture for industry," Information and Software Technology, vol. 193, p. 108039, 2026, doi: 10.1016/j.infsof.2026.108039

    arXiv 2026

Show all 22 references
  1. [9]

    Spectral entropy of early-life distress calls as an iceberg indicator of chicken welfare,

    K. A. Herborn, A. G. McElligott, M. A. Mitchell, V. Sandilands, B. Bradshaw, and L. Asher, "Spectral entropy of early-life distress calls as an iceberg indicator of chicken welfare," Journal of the Royal Society Interface, vol. 17, no. 167, p. 20200086, 2020, doi: 10.1098/rsif...

    2020

  2. [10]

    Systematic review and meta -analysis of thermal stress assessment in poultry using infrared thermography in specific body areas,

    R. C. Hernández-Sánchez et al., "Systematic review and meta -analysis of thermal stress assessment in poultry using infrared thermography in specific body areas," Animals, vol. 14, no. 22, p. 3171, 2024, doi: 10.3390/ani14223171

    2024 doi

  3. [11]

    Optical flow, flock behaviour and chicken welfare,

    M. S. Dawkins, R. Cain, and S. J. Roberts, "Optical flow, flock behaviour and chicken welfare," Animal Behaviour, vol. 84, no. 1, pp. 219–223, 2012, doi: 10.1016/j.anbehav.2012.04.036

    2012 doi

  4. [12]

    AI -powered vocalization analysis in poultry: Systematic review of health, behavior, and welfare monitoring,

    V. Manikandan and S. Neethirajan, "AI -powered vocalization analysis in poultry: Systematic review of health, behavior, and welfare monitoring," Sensors, vol. 25, no. 13, p. 4058, 2025, doi: 10.3390/s25134058

    2025 doi

  5. [13]

    Multimodal digital sensing of early-life laying hens: A pilot study integrating thermal, acoustic, optical - flow and environmental data,

    Y. Dhaliwal, D. Essien, and S. Neethirajan, "Multimodal digital sensing of early-life laying hens: A pilot study integrating thermal, acoustic, optical - flow and environmental data," Frontiers in Veterinary Science, vol. 13, p. 1796102, 2026, doi: 10.3389/fvets.2026.1796102

    2026

  6. [14]

    Multimodal AI systems for enhanced laying hen welfare assessment and productivity optimization,

    D. Essien and S. Neethirajan, "Multimodal AI systems for enhanced laying hen welfare assessment and productivity optimization," Smart Agricultural Technology, vol. 12, p. 101564, 2025, doi: 10.1016/j.atech.2025.101564

    2025

  7. [15]

    Digital twins in agriculture: Orchestration and applications,

    M. Escribà-Gelonch, S. Liang, P. van Schalkwyk, I. Fisk, N. V. D. Long, and V. Hessel, "Digital twins in agriculture: Orchestration and applications," Journal of Agricultural and Food Chemistry, vol. 72, no. 19, pp. 10737–10752, 2024, doi: 10.1021/acs.jafc.4c01934

    2024 doi

  8. [16]

    A multi-layered digital twin framework for smart agriculture and livestock: System architecture and implementation challenges,

    A. Albayrak, "A multi-layered digital twin framework for smart agriculture and livestock: System architecture and implementation challenges," Information and Software Technology , vol. 195, p. 108139, 2026, doi: 10.1016/j.infsof.2026.108139

    2026

  9. [17]

    Harnessing digital twins: Transforming livestock supply chain management: A comprehensive review,

    R. M. Ellahi et al., "Harnessing digital twins: Transforming livestock supply chain management: A comprehensive review," Cogent Food & Agriculture, vol. 12, no. 1, 2026, doi: 10.1080/23311932.2025.2611196

    2026

  10. [18]

    Digital twins in livestock farming,

    S. Neethirajan and B. Kemp, "Digital twins in livestock farming," Animals, vol. 11, no. 4, p. 1008, 2021, doi: 10.3390/ani11041008

    2021 doi

  11. [19]

    State -space models for ecological time -series data: Practical model-fitting,

    K. Newman, R. King, V. Elvira, P. de Valpine, R. S. McCrea, and B. J. T. Morgan, "State -space models for ecological time -series data: Practical model-fitting," Methods in Ecology and Evolution , vol. 14, pp. 26 –42, 2023, doi: 10.1111/2041-210X.13833

    2023 doi

  12. [20]

    State -space models of individual animal movement,

    T. A. Patterson, L. Thomas, C. Wilcox, O. Ovaskainen, and J. Matthiopoulos, "State -space models of individual animal movement," Trends in Ecology & Evolution , vol. 23, no. 2, pp. 87 –94, 2008, doi: 10.1016/j.tree.2007.10.009

    2008 doi

  13. [21]

    Description of a detection model for oestrus and diseases in dairy cattle based on time series analysis combined with a Kalman filter,

    R. M. de Mol, A. Keen, G. H. Kroeze, and J. M. F. H. Achten, "Description of a detection model for oestrus and diseases in dairy cattle based on time series analysis combined with a Kalman filter," Computers and Electronics in Agriculture, vol. 22, no. 2 –3, pp. 171 –185, 1999...

    1999 doi

  14. [22]

    Estimating egg production declines at high environmental temperatures and humidities,

    J. M. Zulovich and J. A. DeShazer, "Estimating egg production declines at high environmental temperatures and humidities," Amer. Soc. Agricult. Eng., ASAE Paper No. 90-4021, 1990. Yashan Dhaliwal (Undergraduate Student Member, IEEE) received the B.Sc. degree in bioveterinary s...

    1990

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.