The Moseley E-MCP is a self-powered hockey-puck-sized device that sits on top of pallets inside refrigerated containers, cold rooms and ripening rooms. It holds an ethylene-action payload bound to a metal-organic-framework matrix, carries an on-board gated gravimetric sensing cavity to read its own headspace, and decides for itself when to release — or accepts orchestration commands from E-Aegis. Forty-two days of unattended operation.
Traditional climacteric-modulator delivery is a one-shot passive sachet or a centralised gas charge — fixed dose, fixed location, no awareness of where the ethylene-producing fruit actually is, no awareness of where the ripening hotspots have moved over time. The E-MCP is the opposite of that: a small autonomous release node, deployed in numbers across the cargo space, releasing payload only when the on-device or upstream model says it is warranted.
Pucks deploy on top of, or alongside, palletised cargo. Spatial coverage replaces the single back-wall dose so headspace concentration tracks the actual ripening front through the load.
Each node runs an on-device TinyML model distilled from E-Array, E-Aegis and cloud teacher models. Release is triggered by the rate of change of ethylene seen by the puck's own gated sensing cavity — or by orchestration commands from E-Aegis when present. Never by absolute ppb thresholds.
The sealed cartridge holds the ethylene-action payload coordinated inside a metal-organic-framework matrix. A heating element beneath the cartridge liberates payload; a rotating-disc shutter and condensation-protected vent dispense vapour into the local headspace.
The receptor chemistries that detect ethylene at trace levels — metal-organic frameworks functionalised with biomimetic ethylene receptors — cannot survive 42 days of continuous reefer humidity. The E-MCP solves this by placing the sensing material inside a sealed cavity that opens only briefly, on a configurable duty cycle. The receptor is protected for the majority of the deployment; ethylene is measured during short, controlled sampling windows.
A small chamber inside the housing carries a MOF functionalised with a biomimetic ethylene receptor. A gate or shutter seals it from the headspace between samples — no condensation, no humidity exposure, no receptor poisoning.
Configurable intervals from one minute to 24 hours; exposure windows from one second to 30 minutes. During exposure, ethylene adsorbs onto the receptor and a gravimetric sensing element — typically a quartz crystal microbalance — measures the mass change.
The on-device inference engine reads adsorption rate, acceleration and recovery profile across successive samples. It infers climacteric progression and emerging excursion conditions — then issues a release decision without any external controller required.
A laboratory-grade ethylene analyser — the kind that resolves single-digit parts-per-billion — is expensive, fragile and impractical to put on every pallet. The Moseley E-Aegis carries one of these instruments and uses it to learn what real climacteric events look like over thousands of cold-chain hours. That trained knowledge is then distilled into a TinyML model that runs on the E-MCP. The result: a hockey-puck-sized release node that infers climacteric progression with lab-grade decisioning, without carrying a lab-grade sensor.
Distillation is a machine-learning technique in which a small, cheap model is trained to imitate the decisions of a large, expensive model. The big “teacher” learns from rich ground-truth data; the small “student” learns from the teacher’s outputs. The student ends up reproducing nearly the same decisions on commodity hardware.
The E-Aegis carries a ppb-accurate ethylene instrument and a multi-sensor array conditioned by E-Conditioning. It records real climacteric events under real cold-chain conditions and labels them — the kind of labelled dataset you can’t get from a spec sheet.
The distilled TinyML model is flashed onto every E-MCP. The puck doesn’t need a ppb-accurate sensor of its own — its gated gravimetric cavity gives it just enough signal, and the inherited model fills in the rest. Lab-grade decisioning at sticker-pack economics.
This is the core economic argument for distributed climacteric control: you only need to pay for the expensive instrument once — at the E-Aegis layer — and that intelligence then cascades to every E-MCP node in the fleet.
From top to bottom: a louvred vent cap protects the airway. A rotating-disc shutter inside a sealed chamber gates vapour egress and isolates the cartridge from condensation through a hydrophobic membrane and condensation trap. The release cartridge holds the ethylene-action payload bound to a metal-organic-framework matrix. A planar heating element beneath the cartridge drives controlled release. The control PCBA carries BLE 5.3, NFC cartridge identity and the on-device inference model. A flat battery pack of puck diameter provides primary energy; an organic-photovoltaic sidewall augments it for the 42-day operational floor. A base seals the assembly.
The architecture is detect → predict → remediate. The E-Nose Label and E-Aegis sit on the detection side. The E-Sentinel ML model and the Metal-Cystine Sensor Array sit on the prediction side. The E-MCP — alongside the E-Remediation layer of the E-Aegis — sits on the remediation side, but as a distributed, networked actuator rather than a single bed inside the sentinel device.
A typical 40′ reefer carries several E-MCP nodes on top of the pallet stacks, an E-Aegis sentinel at the front bulkhead near the reefer return-air path, and per-carton E-Nose Labels in high-value cartons. Reefer return-air carries each puck's controlled vapour egress through the headspace; release events are coordinated so dose is delivered to identified ripening hotspots rather than the load as a whole.
Nodes deploy on top of palletised cargo. Spatial coverage replaces a single fixed-dose source.
The sentinel device coordinates release events and bridges the BLE mesh to the upstream gateway.
Each node's vapour egress is entrained into the container's return-air path so dose reaches the hotspots that matter.
The sensing network logs the headspace response after every release and updates the forecast for the next event.
Reefer container fleets, ripening-room operators and high-value DC managers — we are partnering with cold-chain operators on instrumented trials of the full Moseley Climacteric architecture, of which the E-MCP is the actuator layer.
Discuss a deployment →