How does the smart storage management platform fit?
The difficulty of a smart storage platform often arises after the device has been "connected": the tray passes through, the reader is not in the library, but the WMS does not have a normal flash; the re-establishment of the network leads to a double change in inventory; and the replacement of an equipment model, with changes to the top operating procedures. To avoid these problems, it requires a clear field view, business judgement and a division of effort for implementing feedback, and then selection of protocols and devices. This paper uses the RFID channel as its main link, discussing how a platform can be built to connect, remove barriers and expand.
From the moment a consignment passes through the tunnel, draw the information flow that is really needed.
Write the business: whether the tray has a valid identity, who maintains the box-to-ship relationship, which task allows this operation, and who confirms the change in inventory. Without these conditions, the platform cannot judge it as receipt, exit, temporary storage or return.
It is proposed that the link be structured into five distinct lines of responsibility. They are logical boundaries and do not require that small projects necessarily deploy five separate sets of services.
| Link | Main duties | Results to be delivered |
|---|---|---|
| Labels and Devices | Identification of objects, with readings and the context of the device | Label lists, antennas or points, collection time, equipment status |
| Edge Fit | Connect devices, convert protocols, and execute a trigger and read window policy | Normalize readings, configuration versions, necessary local caches |
| Event service | Combined reading, associated tasks, creating searchable events | Event identity, range of objects, status of treatment and causes of anomalies |
| WMS and related business systems | Verifying documents, privileges, stock status and submission services | Accept, reject, pending review or completed results of operations |
| On-site feedback and mobility | Feedback of results to operators or control systems, recording of alarms | Clear feedback allowing continuation, need for review, suspension, etc. |
WMS (Standing Management System) assumes warehousing operations, and the movement of transmission lines or automation equipment usually requires WCS (Standing Control System), PLC (Programmable Logic Controlr) or equipment control program to work with. Release should be based on the control logic identified by the project, and an uncertified RFID reading cannot be allowed to drive the implementing agency directly. Nor can the platform page " Show success" substitute for on-site implementation feedback.
Device supports the same label, which is not equal to the previous interface being universal
RFID access distinguishes between two compatibility categories: the agreement to identify labels and readers, and the data interface between the reader and the collector. Even if two readers support the same tag interface in the air, their SDK, network commands, state code and configuration may differ. Only frequency bands or brand names are checked when purchasing them, and it is not possible to determine whether the platform can be accessed directly.
It is recommended to keep compatible matrices for each device: Model, Solid, Interface Version, Operating Environment, Connective Method, Available Fields, Time Source, Breakline Restoration and Validised Site. Access programs provide stable data models up, limiting equipment differences to the appropriate level. The remodel still needs to be tested, but should not force the storage document rules to be re-established.
The appropriate layer must check at least the original format and standard values for the label code, equipment and antenna identity, collecting time sources, available signal information, and whether to repeat it after reconnecting. The field provided by a model cannot be considered as having all the necessary devices; when there is no device clock, it must clearly be recorded by edge nodes and their limitations.
Fixed reading and writing equipment is suitable for controlled workspaces or corridors, handheld devices are suitable for mobile inventory and abnormality review. The number is determined by the point and operation. There is no requirement that fixed equipment and handheld equipment be configured in proportion. Metals, liquids, stacking and labeling problems should be verified under real cargo conditions and cannot be solved by simply replacing frequency bands or increasing power extrapolation.
How the reading becomes a re-testable business event
The same label may be read over and over again while the tray is at the door. The original observation record is suitable for a detoxification, and business events should indicate a candidate with context. You can combine trigger signals, read windows, channel maps, directional judgement and task information, group multiple observations into one candidate event, and then submit them to the operational system for validation.
The event model suggests that the following information be clarified and field names may be adjusted to project:
- Event Identity:The same submission and retransmittal of a stable event number, recording the nodes and serial numbers of origin; the next legitimate operation uses new operational identities.
- Observation context:Read points, equipment, antennas, time frames, label collections and whether they have been manually collected.
- Operational context:Warehouses, tasks, documents, object hierarchy and rule versions; not yet matched should be clearly identified.
- Process result:To be matched, received, submitted, rejected or reviewed and to record the status of confirmation and retesting.
The location field is not misused. GS1 EPCIS distinguishes between the headPoint and the bizLocation, and between the time of the event and the time of the system recording. The project can draw on these concepts: Channel reading points do not necessarily equal the post-good repository, nor does the time of receipt equal the time of physical passage.[1]
It is not recommended that “the same EPC keep only one record” be used as the full business data. This covers a status check, but it does not explain what went through the middle, nor can it be easy to track duplicate submission and cancellation operations. Original observations, business events can be kept in layers with current inventory, and different retention periods are configured by business value.
The network program first checks the faulty boundaries and then compares bandwidth.
Fixed-spaced jobs can be prioritized for wired access; mobile terminals need to be linked to actual routes to verify wireless coverage and roaming. Before selecting PoE, it is necessary to check whether the equipment supports the power supply, the electricity standards and the total power budget of the switch, which cannot be equated with "wireable power through the Internet." The serial device also needs a gateway or matching program, RS485 is a physical interface, and does not automatically specify which business agreement is used at the upper level.
The field study should include full load shelves, vehicle passage, terminal movement and simultaneous operating conditions. Besides swallowing, you have to observe the drops, time lags, and repeat times and short breaks. Reading areas interact with the network is unstable, and it is a problem at different levels, not to be mixed up with "signal differentials."
The network design should also specify the address of the equipment to its asset, whether it is actively reported on by the platform or the facility, require open services, time-synchronous approaches and how remote transport will enter. The equipment network, the management network and the business interface should be authorized between them as they are actually needed, so as to avoid sharing default accounts or opening unnecessary services for long periods of time in order to facilitate connection.
The capacity planning should also come from sampling: measure the original reading peaks, merge the event amount, increase batches and logs, and then estimate resources in conjunction with time. Do not use fixed multiple redundancy as a basis for all warehouses ' design, especially when reloading multiple nodes after the offline.
After the net's off, what goes on and what has to be suspended.
The log count is classified by operational nature. Recording an inventory is observed using different offline rules than confirming release; operations involving the validity of a mission and inventory occupancy cannot continue by default because there are old documents in local caches. The range allowed to go offline, the life of the cache, manual confirmation and suspension conditions should be determined before going online.
If the edge node requires a lasting cache, it should keep the event identity, collect time, send the state and rule version, and specify that the disk will be alarmed and processed when the threshold is reached. The cache only exists in memory, which cannot be proved to be restored once restarted; or saves the file, nor does it mean that the business judgement after retransmission is correct.
Example of re-routing test:After the tray is broken through, the event remains at the edge; the original task was cancelled during the period. Once the connection is restored, the platform needs to recheck the event and mission status, and turn the conflict into a clear result, rather than an automatic one-by-one account of all caches. If the inventory is actually submitted, it is simply returned lost, and the result should be found when retransmitted, so that it does not recur.
Using message communication protocols or lines of messages such as MQTT, you can separate the collection and operations sides for data processing, but not all items must be introduced. In OASIS MQTT, QoS (quality service level)1 allows at least one delivery to enable receiving a duplicate message; even using QoS 2, it is not possible to conclude that WMS database services are "just one execution" on this basis. The transmission and business submission are two boundaries, and still require end-to-end event identity, and other handling and back-to-back mechanisms, i.e., inventory does not change over time when the same business event is retransmitted.[2]
Small-scale, low-mixed projects can use direct interfaces with durable records and retesting; multi-equipment, multiple subscribers or more frequent backups, then evaluate the message channel. The round of inquiry is not natural backward, but should be used in accordance with equipment capacity and time limits. The key is to explain how failure is restored, not whether technical terms are complete.
Platform alarms must be able to answer: What part of the malfunction?
“Equip Online” only indicates that the connection or heartbeat meets the test criteria, does not justify proper antenna reading, labeling and, even less, stock updating.
- Connection layer:Equipment is broken up, frequently reconnected, time abnormal, configuration version not in line.
- Collection level:Triggered but no target reading, abnormal label collections, and capture windows not closed.
- Event Layer:Backlog to be sent, increased number of re-tests, repeat or disorderly events, cache capacity close to threshold.
- Business level:No matching of tasks, conflict of status, denial of interfaces, successful inventory submission but no confirmation of receipt.
For example, "Tray passed without a library" is sorted in order to trigger whether or not it occurs, read whether there is a candidate event, whether the task is matched, whether the inventory is submitted, whether the feedback is delivered. There is time and status on each level, so there is no need for an equipment provider, network personnel and software people to speculate over and over again.
Line order: First, you can put a back-to-back link in.
Select a position with a clear boundary, identify the target and expected business results; complete real cargo identification, model events, interface submission and on-site feedback. Then set up line by line, area tags, grid disconnect reconnection, restarting relays, duplicate submissions and job cancellations, record processing results.
The pilot delivery should not be limited to a bitmap and login number, but should also have equipment compatible matrix, configuration baseline, event field description, business state flow, malfunction recovery manual and test evidence. The configuration can be modified to return to the baseline, historical events can be traced back by number, new devices can be accessed along the already suitable boundary, then extended to multiple corridors and storage areas, so that the platform has a basis for continuous maintenance.
Technical reference
The following criteria are used to check the semantics of events and message delivery boundaries, and the platform need not introduce all standard components for the purpose of quoting these concepts.
- GS1: EPCIS data modelSee eventstime, recordtime, readPoint, bizLocation.
- OASIS:MQTT Version 5.0See the level of service and information delivery process.

Ms. Lei
Mr. Pang