Predictive Waste Collection and Resource Recovery: Research Protocol

Material-account reference — 6 November 2026

A separate offline reference reconciles opening inventory, incoming material, other declared inflows, material and residue outputs, other declared outflows and closing inventory. Four invented accounts demonstrate balanced, incomplete, within-tolerance and unreconciled states. Quantities use integer grams; no recovery percentage is produced. Synthetic software tests check input rules and arithmetic only, not instruments, physical outcomes or downstream recycling.

• Contract, reproduction and interpretation limits
• Offline material-account reference
• Four fictional aggregate accounts
• Synthetic reconciliation counterexamples

The facility Design study describes the proposed wider review. Actual facility records, stock and instrument validation, justified tolerances, downstream evidence and policy comparisons remain dependencies. All earlier preparation files are preserved.

Observation evidence reference — 5 November 2026

A separate offline reference classifies fictional fill observations as missing, not yet received, unknown time, suspect, stale or current at an explicit decision cutoff. It preserves zero as a value and separates observation age from recent arrival. The fifteen-minute limit is an invented fixture parameter. Synthetic software tests establish classification behavior only, not sensor accuracy, operational effectiveness, or policy performance.

• Contract, scope and reproduction commands
• Offline observation classification reference
• Six fictional records and expected states
• Synthetic counterexample tests

The review-queue Design study describes the proposed wider workflow. No interface, sensor network, model or policy experiment is executed by this reference. All earlier collection-ledger files and counting rules remain unchanged.

Research question

How do fixed, reactive, and predictive collection policies compare when waste generation, vehicle availability, segregation quality, and facility constraints change together? The proposed study will test those relationships inside a clearly labelled synthetic operating model.

System boundary

The model will connect scheduled service, collection events, vehicle trips, receiving facilities, sorting, recovered material, and residual disposal. It will distinguish event time from record arrival time and retain explicit states for missing or uncertain observations.

UNEP and ISWA's Global Waste Management Outlook supplies context on waste management and resource recovery. The simulation's assumptions require separate documentation and review; they are not observations extracted from that report. [1]

Planned comparisons

Policies will run against identical seeded scenarios and initial conditions. Stress cases will include breakdowns, changes in demand, contamination, delayed uploads, and restricted facility capacity. The predictive policy may use only information available when each decision is made.

Planned measures include service coverage, unserved demand, overflow duration, travel distance, collected mass, recovered output, rejection, and computation time. Units, denominators, mass-balance rules, and handling of missing data will be fixed before comparing results.

Validity and field transition

Repeated trials and sensitivity analysis will examine dependence on synthetic assumptions. A simulation improvement will describe performance within that model. Field effectiveness will require authorised operational data, calibration, and a supervised evaluation with a suitable baseline.

Release requirements

Any later results must accompany the generator version, seeds, parameter settings, policy code, experiment logs, failure analysis, limitations, and contributor disclosures. The associated design study explains the proposed operational decisions. Neither page reports a completed experiment or a government deployment.

Collection-metric preparation — 29 October 2026

A first preparation pack defines a recorded-completion metric for a fixed synthetic service plan. It includes eight fictional visits, versioned event fixtures, an offline reference function, and a guide. Local software tests exercise duplicate handling, reporting cutoffs, corrections, invalid inputs, and denominator consistency. These are metric-logic checks, not field observations or digital-twin policy experiments.

• Guide, boundaries, and reproduction instructions
• Metric contract and expected fixture snapshots
• Fictional service plan and report events
• Offline Python reference function

The collection ledger design study explains the measurement choices. Field validation, mass balance, vehicle and facility constraints, policy comparisons, and independent technical review remain pending.