Project · Drone & inventory
Drone inventory: seeing the warehouse from above.
In January 2026 I shared a pilot designed to bring a new perspective to inventory work: the drone enters the aisles, captures images and turns a complex check into evidence that can be read and reconciled. Mobiix supported the project with expertise, responsiveness and practical attention to every step.
The priority scope covers approximately 30,000 pallets out of roughly 70,000 in the facility: the first group is the ABC-movement focus, while configuration continues to extend coverage across the warehouse. The approximately 277 hours are an equivalent estimate obtained by applying the previous reference pace to the larger scope; they are not mission time logged in the field.
01 / The initial question
In a high-density warehouse, checking physical locations against system records takes time and requires visibility of levels that are difficult to inspect from the floor. The project explored a drone as a tool for inventory verification and visual evidence collection.
The project presentation identifies a DJI Mavic 3 Enterprise platform. The initial ambition was to automate part of the checks and make acquisition repeatable. This describes the objective, not proof of a fully autonomous or real-time inventory solution.
02 / The work before take-off
My contribution focuses on connecting operational processes, information requirements and stakeholder coordination. The start-up materials document the pilot scope, preparation of rack mappings and comparison of captured data with the warehouse-system reference.
The geometric model needs columns, levels, clear heights and beam dimensions. The sources use millimetres, centimetres and metres, making unit normalisation essential. Missing identifiers, duplicates and inconsistent sequences can undermine the association between an image and its location.
Real racks are not uniformly regular. The documentation covers fire-service access gaps, different level heights and first bays that differ from the rest of the row. These exceptions belong in the configuration: copying a standard geometry without checking the physical layout is insufficient.
03 / From capture to reconciliation
The designed workflow includes area mapping, an inventory reference extract, image and barcode capture, result transfer and reconciliation with the warehouse system. The specifications also include operator selections and commands. Automation must therefore be assessed across the entire process, not flight alone.
Reading a barcode does not establish the quantity or contents of a pallet. Pallet identification, location association and content checking are separate questions. A missing read must remain an exception to investigate, not automatically become an empty location.
A reliable comparison should retain acquisition timestamps and use a consistent warehouse-system snapshot. Movements between extraction, flight and analysis need reconciliation; otherwise timing differences can appear to be inventory errors. These are control requirements, not capabilities claimed here as delivered.
04 / What the images show
The photographs show hanging stretch film, irregular wrapping and uneven visibility. They help direct targeted checks, but do not by themselves demonstrate structural rack damage or stock discrepancies.
Visual inspection has value separate from inventory counting. It documents exceptions for discussion with operations; classification and any intervention remain subject to on-site checks and company procedures.
05 / Beyond inventory
The pilot opened a wider perspective. The same tool can support facility teams with maintenance, damage assessment and inspections of areas that would otherwise require a platform to reach. Bringing a camera to a difficult or intermittent access point makes the first observation easier and gives follow-up checks a shared visual reference.
The point is not to replace an on-site inspection. It is to arrive better prepared, with a traceable visual record and a clearer priority for the next action. That is a practical connection between safety, operational continuity and data quality.
06 / The next frontier: reading video with AI
We are now looking at the next step: using artificial intelligence to read video. The aim is to understand whether images can help distinguish defective, unsafe or damaged pallets and create a prioritised queue for human verification.
The same logic could help prepare the environment for AGVs by highlighting obstacles, unusual conditions and situations to check before extending automation. This is a direction under study, not a feature claimed as completed. Stay tuned.
Evidence from the field
The project in motion
Excerpts from the original materials. Selected framing and identifying details redacted; audio and metadata removed. Photographs document visual observations, not acceptance-test results.
07 / The numbers and what they mean
| Metric | Value | Calculation / limitation |
|---|---|---|
| Priority scope | ≈30,000 pallets | ABC-movement pallets at the centre of the pilot; not a count of successfully identified pallets. |
| Total facility | ≈70,000 pallets | Pallets present in the facility; configuration is being extended to cover the wider warehouse. |
| Equivalent effort | ≈277 hours | 30,000 ÷ 13,000 × 120 hours, keeping the reference pace; not logged flight time. |
| Derived average pace | ≈108 pallets/hour | 30,000 ÷ 277; indicative average, not a constant rate. |
| Battery | 20–30 minutes | Reported operating range, dependent on pilot handling; not DJI’s nominal endurance. |
08 / Benefits and limitations
Benefit · ABC priority
Focusing on the approximately 30,000 pallets in ABC movement gave the pilot a concrete operating scope linked to the positions most relevant to flow.
Benefit · visibility at height
The drone brought a camera close to levels and points that are difficult to observe from the floor, creating visual evidence for inventory and inspections without turning every check into a platform activity.
Benefit · digital foundation
Mappings, images and reconcilable results created a basis for repeatable checks and for extending the configuration to the rest of the warehouse.
Benefit · new applications
The same visual capture can support facility work, maintenance, damage assessment and future AI and AGV studies, broadening the project beyond inventory alone.
Endurance and interruptions
Replacing a battery every 20–30 minutes requires changeovers, charging management and session restarts. The range also depends on pilot handling and cannot alone establish the total number of batteries used during the campaign.
A continuously committed resource
The reported process occupies one person throughout. The drone does not eliminate labour: it shifts work to piloting, supervision and acquisition management. The 120 hours exclude any additional support personnel, whose effort has not been quantified.
Coverage lags behind change
Fifteen working days for the stated scope means that early observations may already be outdated when acquisition ends. Subsequent analysis risks describing a warehouse situation that has already changed.
Data quality and downstream work
Reflections, obscured labels, wrapping and non-uniform geometry call for verification. Without complete logs, read rates, rework, time distributions and final accuracy cannot be quantified. Accuracy targets in planning documents are not measured outcomes.
A flight can change the way we observe.
The project confirmed for me that innovation is not only about introducing technology; it is about connecting images, process and decision. I am grateful to Mobiix for the practical, collaborative support throughout the journey. The inventory pilot is a first chapter in a broader use of drones: more agile inspections, more readable evidence and new questions to take back into the operation.
Confidentiality
Prices, costs and confidential information are not published.