A difficult question.
A useful experiment.

Source freshness and decision quality

How should an interface distinguish observation time, receipt time, confidence, and connection state? What helps a reviewer recognise stale or incomplete evidence?

Disconnected operation and reconciliation

Which tasks can continue with saved context? How should queued reports and command updates reconcile after a device reconnects, without presenting old information as current?

Explainable routing and human authority

How can a system expose changed constraints and tradeoffs clearly enough for an accountable person to review a proposed route?

A useful evaluation starts with an identified software build, devices, network topology, scenario data, and acceptance criteria. Separate external-uplink loss from device-to-edge disconnection. Record the information actually available to each participant.

Possible measures include completion of permitted tasks, stale-information recognition, reconciliation correctness, time to a reviewed decision, and the completeness of its audit record. Baselines, repeated runs, failures, and limitations belong in the same report.

This is a proposed evaluation direction. Test protocols, available resources, and publication rights would be agreed with a collaborator before work begins.

Realvi’s public website scenario uses a fictional 14-node, 16-road graph. It illustrates route constraints, connection states, and explicit approval. The engineering note explains what is calculated and the limits of the demonstration.

The scenario is not the production platform, a benchmark against other products, or evidence of a completed operational trial. Actual prototype and lab evaluations require their own methods and results.

Potential formats include a scoped technical review, a supervised student project through an institution’s normal process, a joint study, or a laboratory evaluation. A useful first conversation identifies the question, contribution from each side, required resources, and intended output.

Relevant expertise includes distributed systems, logistics optimisation, human factors, decision support, and software testing. We are open to critique that helps the product become more useful.