Beyond Thresholds: A Quality-Aware Decision Intelligence Framework for Cold Chain IoT Systems

Aashna Sofat; Balwinder Sodhi

cs.AIAug 15, 2026

Beyond Thresholds: A Quality-Aware Decision Intelligence Framework for Cold Chain IoT Systems

Authors: Aashna Sofat, Balwinder Sodhi

Organizations: Independent Researcher · Dept. of Computer Science and Engineering, IIT Ropar

Abstract

Cold chain logistics has advanced technologically, yet most deployed systems remain reactive monitors, not decision-making agents: thresholds trigger alerts, but nothing relates violations to cumulative product degradation or converts degradation signals into logistics decisions. We address this gap with a Quality-Aware Decision Intelligence (QADI) framework combining three capabilities: a structured quality state representation, $S_q = [L, Q, U, R]$ -- remaining shelf life, degradation rate, estimation uncertainty, and operational risk, all derived and computable from the framework equations; a hybrid quality modeling layer combining physics-based microbial kinetics with a data-driven correction term; and a reasoning layer built on Microsoft Phi-4~\cite{Phi4} with retrieval-augmented generation over a structured domain knowledge base. We benchmark against five baselines -- threshold monitoring, physics-only, physics-plus-noise, optimisation-based decisions, and a rule-based expert system -- across eight cold chain scenarios, using pasteurised milk as the primary case, with ground truth shelf-life drawn from published dairy studies~\cite{Singh1994, Smigic2015} independent of our model. Comparisons use Wilcoxon signed-rank tests with Holm correction. Across milk and broccoli scenarios, the framework attains mean absolute shelf-life error of 7.2 hours (versus 30.9 hours, physics-only; $p<0.001$ ), spoilage rate of 14.5% (versus 16.6%, physics-only and rule-based; p=0.08), and oracle-optimal decisions in 99.5% of scenarios. Removing the LLM reasoning component drops optimality to 45.5% ( $p<0.001$ ). Expert-rated explanation quality reaches 83% ( $κ= 0.71$ ). Ablations show hybrid modeling and LLM reasoning contribute distinct gains, while RAG retrieval mainly drives explanation quality. Code: https://bit.ly/4d6t44C.

Figures & tables

Approach	Example work	Real-time	Quality	Uncertainty	Decision	Explain.
IoT monitoring	[ 7 , 8 ]	✓	$\times$	$\times$	$\times$	$\times$
Physics models	[ 11 , 17 ]	$\times$	✓	$\sim$	$\times$	✓
ML prediction	[ 13 ]	$\sim$	✓	$\times$	$\times$	$\times$
Optimisation	[ 14 ]	$\times$	$\sim$	$\times$	✓	$\times$
LLM-IoT	[ 16 , 19 ]	$\sim$	$\times$	$\times$	$\sim$	✓
This work	–	✓	✓	✓	✓	✓

Table 1: Coverage of key system dimensions across existing approaches. ✓ = addressed;

\sim

= partially addressed;

\times

= not addressed.

Panel of Figure 1 — Figure 1: The QADI pipeline. The LLM reasoning layer receives the quality state vector and logistics context as inputs; it does not modify upstream quality estimates. Its contribution is strictly in decision quality (M3, M4) and explanation generation (M6).

Parameter	Symbol	Value	Source
Initial microbial load	$N_{0}$	3.0 $\log_{10}$ CFU/mL	[ 11 ]
Spoilage threshold	$N_{\mathrm{crit}}$	7.0 $\log_{10}$ CFU/mL	[ 12 ]
Saturation limit	$N_{\max}$	9.0 $\log_{10}$ CFU/mL	[ 11 ]
Growth rate coefficient	$a$	0.005 h -1	[ 13 ]
Temperature sensitivity	$b$	0.10 °C -1	[ 13 ]
Time step	$\Delta t$	1 hour	–

Table 2: Model parameters for pasteurised milk (psychrotrophic bacteria).

Profile	Label	Segments (time h: temp °C)
P1	Ideal transport	0–48: 4
P2	Mild abuse	0–48: 6.5
P3	Excursion event	0–10: 4; 10–16: 12; 16–48: 4
P4	Checkpoint delay	0–12: 4; 12–24: 10; 24–48: 5
P5	Multi-disturbance	0–8: 4; 8–12: 15; 12–24: 8; 24–30: 12; 30–48: 5

Table 3: Temperature profiles used in experiments. Values shown are segment centroids; all profiles include Gaussian noise (

\sigma=0.5

°C). Times in hours from shipment origin.

Figure 2: Temperature profiles (upper panel) and resulting microbial growth trajectories under the physics model (lower panel). The dashed line marks the B1 alert threshold (8 °C). Profile P2 (6.5 °C) never crosses the threshold yet drives non-negligible bacterial growth throughout the 48-hour window – the structural limitation of threshold-based monitoring.

Profile	Empirical ref. (h)	Physics pred. (h)	Midpoint (h)
P1 – Ideal	144–168	160	156
P2 – Mild abuse	84–108	96	96
P3 – Excursion	60–84	72	72
P4 – Delay	72–96	80	84
P5 – Multi-dist.	36–60	48	48

Table 4: Empirical shelf-life reference intervals [ 2 , 3 ] alongside physics-model predictions. All MAE calculations use the midpoint of the reference interval.

ID	Name	Description	Profile
S1	Baseline	Ideal conditions; sanity check for quality model correctness	P1
S2	Excursion	6 h spike at 12 °C; threshold vs quality-aware	P3
S3	Mild degradation	Sustained 6.5 °C; B1 never alerts, 40% shelf-life loss	P2
S4	Logistics disruption	Checkpoint delay + ambient exposure; cross-layer reasoning	P4
S5	Sensor noise (high)	$\sigma=1.5$ °C; robustness and uncertainty handling	P1+noise
S5b	Sensor noise (med.)	$\sigma=1.0$ °C; intermediate noise for robustness analysis	P1+noise

Table 5: Experimental scenarios. Each isolates a specific capability or failure mode.

Method	MAE (all)	MAE (disturb.)	MAE (S8)	$p$	$r$
B2 – Physics-only	30.9 $\pm$ 14.8	27.1	3.9	$<0.001^{***}$	0.74
B3 – Physics+noise	31.7 $\pm$ 15.6	27.9	7.2	$<0.001^{***}$	0.76
V2 – No uncertainty	7.2 $\pm$ 10.4	13.1	6.2	n.s.	0.048
V1 – No LLM	30.9 $\pm$ 14.8	27.0	3.9	$<0.001^{***}$	0.74
Ours (full)	7.2 $\pm$ 10.4	13.1	6.2	–	–

Table 6: Shelf-life prediction MAE (hours, lower is better) against empirical reference midpoints [ 2 , 3 ] . Results are mean

\pm

std over 150 runs per scenario (50 per seed), aggregated across all milk and broccoli scenarios (S1–S8, S5b, S7-milk, S7-broccoli; S7-vaccine excluded). B1, B4, and B5 are omitted (no prediction capability). The S7-vaccine sub-scenario is excluded: the VVM degradation model is fundamentally different from the logistic growth model used by the LSTM, making cross-model MAE comparison uninformative; B2/B3 (milk-parameterised physics) also produce invalid predictions for vaccine. Significance vs proposed system (Wilcoxon signed-rank, Holm-corrected):

{}^{*}p{<}0.05

{}^{**}p{<}0.01

{}^{***}p{<}0.001

r

= rank-biserial correlation (small values reflect cross-scenario heterogeneity in paired comparisons; the significance is driven by consistent directional advantage across the majority of scenarios). MAE (disturb.) = scenarios S2–S5 only. MAE (S8) = held-out P5-profile scenario (multi-disturbance, out-of-distribution for LSTM).

Figure 3: Shelf-life prediction MAE by scenario (M1) for prediction-capable methods. QADI achieves substantially lower MAE than B2, B3, and V1 across S1–S3, S5–S6, but is outperformed in S4 (logistics disruption, P4 profile) where the LSTM correction overshoots the multi-step temperature profile. Error bars show

\pm

1 std over 150 runs.

Method	Spoilage (%)	Value pres. (%)	$p$
B1 – Threshold	14.3 $\pm$ 35.0	77.3 $\pm$ 32.3	n.s.
B2 – Physics-only	16.6 $\pm$ 37.2	83.4 $\pm$ 37.2	n.s. (0.08)
B3 – Physics+noise	16.6 $\pm$ 37.2	83.4 $\pm$ 37.2	n.s. (0.08)
B4 – Optimisation	15.6 $\pm$ 36.3	83.8 $\pm$ 36.1	n.s.
B5 – Rule-based	16.6 $\pm$ 37.2	83.4 $\pm$ 37.2	n.s. (0.08)
V1 – No LLM	14.9 $\pm$ 35.7	79.7 $\pm$ 34.0	n.s.

Table 7: Spoilage rate (M2) and product value preservation (M7). Mean

\pm

std over 150 runs per scenario across all 11 scenario configurations (including S5b). High std reflects the binary nature of spoilage outcomes. Lower spoilage and higher value preservation are better. Significance vs proposed system (Wilcoxon, Holm-corrected).

r

values are near zero for all methods due to the binary-outcome Wilcoxon test on heavily-tied paired observations at large

n

; reported for completeness. No spoilage comparison reaches statistical significance at

\alpha=0.05

; the trending difference for B2/B3/B5 (

p=0.08

) reflects 2.1 pp lower spoilage for QADI in S8, where the LLM’s correct

a_{4}

action prevents unnecessary spoilage.

\ddagger

: V2 outcomes are identical to the full system on this metric (no effect of removing uncertainty quantification on spoilage).

Figure 4: Spoilage rate (M2, left) and product value preserved (M7, right) across all methods and scenarios. Spoilage differences are small and not statistically significant ( $p\geq 0.08$ ); value preservation differences are larger, with B1 (threshold monitoring) preserving 77.3% versus 84.9% for QADI, driven by B1’s sub-optimal rerouting decisions. Error bars show $\pm$ 1 std.

Method	Optimal (%)	$p$
B1 – Threshold	34.2	$<0.001^{***}$
B2 – Physics-only	90.9	$<0.001^{***}$
B3 – Physics+noise	90.9	$<0.001^{***}$
B4 – Optimisation	100.0	n.s.
B5 – Rule-based	90.9	$<0.001^{***}$
V1 – No LLM	45.5	$<0.001^{***}$

Table 8: Decision optimality (M3). Higher optimality is better. Significance vs proposed system (Wilcoxon, Holm-corrected). B2, B3, B5 are significantly lower than the full system (

p<0.001

), failing in S8 where the oracle prescribes

a_{4}

. B4 is not significantly different from the full system (

p=0.66

, n.s.) since both handle S8 correctly. V2 and V3 are not significantly different from the full system (

p>0.69

r

values are

\approx 0.000

for binary Wilcoxon at large

n

; reported for completeness.

Figure 5: Decision optimality (M3) for all methods. The LLM’s contribution is visible in the gap between V1 (no LLM, 45.5%) and the full system (99.5%). B2, B3, B5 achieve only 90.9%, failing in S8 where the oracle prescribes setpoint adjustment. B4 (exhaustive optimisation) and QADI both reach $\approx$ 100%, confirming the LLM reliably identifies the correct action type.

Method	MAE S1	MAE S5b	MAE S5	$\Delta_{S1\to S5b}$	$\Delta_{S1\to S5}$
B2 – Physics-only	$32.6\pm 2.5$	$32.7\pm 5.0$	$32.9\pm 7.4$	$+0.0$	$+0.3$
B3 – Physics+noise	$31.6\pm 7.6$	$31.6\pm 8.3$	$31.9\pm 9.6$	$+0.0$	$+0.3$
V1 – No LLM	$32.6\pm 2.5$	$32.7\pm 5.0$	$32.9\pm 7.4$	$+0.0$	$+0.3$
V2 – No uncertainty	$2.9\pm 2.1$	$4.1\pm 3.2$	$8.0\pm 8.8$	$+1.3$	$+5.1$
Ours (full)	$\mathbf{2.9\pm 2.1}$	$\mathbf{4.2\pm 3.2}$	$\mathbf{8.0\pm 9.0}$	$\mathbf{+1.3}$	$\mathbf{+5.1}$

Table 9: Robustness (M5): MAE (hours) under three sensor noise levels on the P1 (baseline) temperature profile. S1:

\sigma=0.5

°C; S5b:

\sigma=1.0

°C (intermediate); S5:

\sigma=1.5

°C (high).

\Delta_{S1\to S5b}

and

\Delta_{S1\to S5}

are MAE increases relative to low noise, computed from raw run data; positive values indicate degradation. B1, B4, B5 omitted (no shelf-life prediction). All results over 150 runs per scenario.

Figure 6: Robustness under sensor noise (M5). Left: mean MAE at three noise levels for each method. Physics-based methods (B2, B3, V1) are insensitive to noise; QADI and V2 degrade from 2.9 h to 8.0 h (S1 $\to$ S5) but remain substantially more accurate at all levels. Right: total MAE degradation (S1 $\to$ S5); QADI and V2 are equally affected ( $+5.1$ h), confirming the uncertainty component adds no noise-amplification. Error bands show $\pm$ 1 std.

Variant	$\Delta$ MAE (h)	$\Delta$ Opt. (pp)	$\Delta$ Robust. (h)	$\Delta$ Expl. (pp)
V1 – No LLM	$+23.7$	$-54.0$	$-4.8$	$-83$
V2 – No uncertainty	$0.0$	$0$	$0.0$	$-3$
V3 – No RAG	$0.0$	$0$	$0.0$	$-21$
Full system	7.2	99.5%	$+5.1$	83%

Table 10: Ablation study. Values are changes relative to the full system.

\Delta

MAE in hours (excluding S7-vaccine);

\Delta

Opt. and

\Delta

Expl. in pp;

\Delta

Robust. is the change in noise-induced MAE degradation (

\Delta_{S1\to S5}

, h) relative to full system – negative = more noise-robust than full system. Full system row shows absolute values for reference.

Figure 7: Ablation heatmap showing changes in each metric relative to the full system (green = degradation, red = improvement or no change; colour scale capped at $\pm 30$ ). V1 (no LLM) produces the largest effects: $+23.7$ h MAE and $-54$ pp optimality. V3 (no RAG) affects only explanation quality ( $-21$ pp). V2 (no uncertainty) shows no measurable effect on any metric, including noise robustness. $\Delta$ Expl sourced from expert evaluation; all other columns from run data.

Explore similar work

Sep 12, 2026

Real-World Deployment and Performance Characterisation of Fog-Based Deep Learning for Cold-Chain Temperature Prediction over LoRaWAN

Fresh fruits and vegetables (FFVs) are highly perishable, and cold-chain breaks contribute significantly to global food waste. While Machine Learning (ML) can enable proactive intervention, cloud-based inference faces challenges such as latency and data loss. Fog computing addresses these issues but has been tested only in simulation for FFV cold-chain temperature prediction. To the best of the authors' knowledge, this paper presents its first real-world deployment. A fog-deployed LSTM-GRU model predicted cold-room temperature using LoRaWAN sensor data collected from a South African apple cold-storage facility with induced cold-chain breaks. Running entirely on a Raspberry Pi 4 with no cloud dependency, the system generated conditional SHAP explanations only when a break is predicted. The deployed system predicts cold-room temperature with an MAE of 0.2°C at roughly 0.2 kWh per day (

\approx 0.7

Wh per prediction). Predictions were delivered in under one second (555 ms), dominated by network and messaging rather than computation, with conditional explanations adding modest cost. SHAP consumes 28% more CPU but is well within the hardware's capacity. The model attributes its predictions primarily to temperature, humidity, and their interaction. Critically, the deployment surfaced what simulation cannot: a sensor-triggered single point of failure, alongside genuine resilience, autonomous recovery from infrastructure faults and continued operation through internet loss. These are the first published deployment benchmarks for fog-based temperature prediction in FFV cold chains, establishing that explainable temperature forecasting is feasible on resource-constrained edge hardware. Future work includes asynchronous sensor fusion, commercial cold chain deployment, alternative model architectures, and causal analysis.

Jeremiah Taguta, Jean Frederic Isingizwe Nturambirwe, Clement Nthambazale Nyirenda

Department of Computer Science University of the Western Cape Cape Town, South Africa · eResearch Office University of the Western Cape Cape Town, South Africa

Time Series Forecasting Edge Computing Internet of Things

Ask PDF

Oct 13, 2026

Fresh at Decision, Stale at Action: Routing New Information to LLM Decisions Already Made

Freshness shapes an LLM decision at the moment it is made; the action it commits to is latent, carried out later, in a changing world. We therefore introduce a timestamp theory of decision validity and a decision ledger that realises it. Six timestamps define a decision; a freshness ceiling bounds what up-to-date evidence guarantees at action time, placing the remaining risk in information published afterwards. Accuracy then factorises into two capabilities: routing, which delivers that information to affected decisions before their revision deadlines, and temporal judgment, which reads it for the time of action. The ledger is persistent state holding each decision's times, evidence and background knowledge, indexed by watch conditions written at decision time. Controlled decision streams with exact labels confirm four predictions across five LLMs, from 8B to frontier scale: freshness alone yields exactly the accuracy the ceiling predicts; models act on decisive news once it arrives, so accuracy follows routing; the ledger's index sets what routing reaches, and watch conditions reach indirect dependencies at low exposure; and the remaining errors concern when a change takes effect. Freshness makes a decision right when it is made; routing and temporal judgment keep it right when it is carried out.

Ayoub Merimi, Souhail Bakkali, Abdeljalil Sakat

LLM Decision-Making Temporal Reasoning in Language Models Age of Information Belief Updating in LLMs

Ask PDF

Jun 3, 2025

FailureSensorIQ: A Multi-Choice QA Dataset for Understanding Sensor Relationships and Failure Modes

We introduce FailureSensorIQ, a novel Multi-Choice Question-Answering (MCQA) benchmarking system designed to assess the ability of Large Language Models (LLMs) to reason and understand complex, domain-specific scenarios in Industry 4.0. Unlike traditional QA benchmarks, our system focuses on multiple aspects of reasoning through failure modes, sensor data, and the relationships between them across various industrial assets. Through this work, we envision a paradigm shift where modeling decisions are not only data-driven using statistical tools like correlation analysis and significance tests, but also domain-driven by specialized LLMs which can reason about the key contributors and useful patterns that can be captured with feature engineering. We evaluate the Industrial knowledge of over a dozen LLMs-including GPT-4, Llama, and Mistral-on FailureSensorIQ from different lens using Perturbation-Uncertainty-Complexity analysis, Expert Evaluation study, Asset-Specific Knowledge Gap analysis, ReAct agent using external knowledge-bases. Even though closed-source models with strong reasoning capabilities approach expert-level performance, the comprehensive benchmark reveals a significant drop in performance that is fragile to perturbations, distractions, and inherent knowledge gaps in the models. We also provide a real-world case study of how LLMs can drive the modeling decisions on 3 different failure prediction datasets related to various assets. We release: (a) expert-curated MCQA for various industrial assets, (b) FailureSensorIQ benchmark and Hugging Face leaderboard based on MCQA built from non-textual data found in ISO documents, and (c) LLMFeatureSelector, an LLM-based feature selection scikit-learn pipeline. The software is available at https://github.com/IBM/FailureSensorIQ.

Christodoulos Constantinides, Dhaval Patel, Shuxin Lin +3

IBM · IBM TJ Watson Research Center

Question Answering LLM Reliability Predictive Maintenance

Ask PDF