DJI Dock 3 · Autonomous drone station
Autonomous drones for railway infrastructure monitoring
Continuous diagnostics.
A distributed fleet of autonomous drone stations — Dock 3 and Matrice 4TD — with on-board artificial intelligence, coordinated by a remote operations centre. From periodic inspection to continuous, predictive, traceable monitoring.
Six structural constraints of inspection work today.
Manual inspection has clear limits: personnel exposure, discontinuous coverage, slow reaction times, data that cannot be compared. These are the problems an autonomous UAV with on-board AI solves.
Exposed workforce
Travelling teams covering long distances, exposed to worksite risks and to the proximity of live track.
Discontinuous coverage
Low-frequency scheduled inspections, on-demand checks triggered only after the event.
Slow reaction times
Slow post-event mobilisation (landslide, flooding, vandalism); narrow intervention windows.
Heterogeneous data
Surveys that cannot be compared, poor historical traceability, difficult change detection over time.
Reactive vegetation management
Highly variable seasonal growth, reactive rather than predictive cutting.
Personnel safety
Work in cuttings, tunnels and on viaducts: reducing human exposure is a strategic objective.
Everything an autonomous drone can replace or enhance.
Nine families of inspection activities, from linear ones on the track superstructure and the overhead contact line through to 3D surveying, thermography and recognition AI. All carried out from the ground, with georeferenced, comparable data ready for automatic ticketing.
Track and rail inspection
High-resolution visual inspection of turnouts, switches, joints and fastenings. Detection of geometric and wear anomalies.
Vegetation and right-of-way
Vegetation growth mapping, identification of clearance-gauge encroachments, predictive planning of cutting.
Topographic survey
High-accuracy DTM/DSM for worksites, alignment changes, slope and drainage analysis.
3D survey and digital twin
Photogrammetric models of stations, bridges and tunnels. The basis for BIM and the asset's digital twin.
Overhead contact line (OCL)
Inspection of masts, cantilevers, insulators and catenary. Thermography for hot spots and defect identification.
Civil structures
Bridges, viaducts, tunnels, retaining walls: visual and thermal inspection, detection of cracks and water ingress.
Security and intrusion prevention
Patrolling of yards, depots and sensitive areas. Alarm verification and deterrence in coordination with CCTV.
Rapid post-event response
Updated survey after a landslide, flood, accident or act of vandalism. Data in minutes, not days.
Recognition AI
Automatic classification of defects, components and obstacles. Georeferenced output ready for ticketing.
An autonomous station in real operation.
The end-to-end workflow: station deployment, automatic take-off, inspection mission along the railway right-of-way, landing and autonomous recharging. The pilot never leaves the operations centre.
Operational demo · DJI Dock 3 + Matrice 4TD
Dock 3 + Matrice 4TD. The station, the drone, the system.
The Dock 3 is the first DJI station that can also be installed on a vehicle: rapid deployment at a mobile worksite or a fixed position, IP56 protection, 24/7 operation with no on-site personnel. It houses the Matrice 4TD — an industrial multi-sensor platform with radiometric thermography, tele/wide cameras, measuring laser and integrated RTK.
DJI Dock 3
Autonomous drone station, fixed or vehicle-mounted.
Matrice 4TD
Industrial multi-sensor platform for diagnostic missions.
AI on board and at the dock. No cloud, no latency.
True autonomy is not automatic take-off — it is the system's ability to understand and decide locally. Our architecture combines a compute pack on board the drone with an edge module at the dock, ensuring real-time detection, data compliance and independence from the cloud.
Computing power equivalent to compact servers, in just 120 grams, mounted directly on the drone. In-flight inference for minimum-latency detection, gimbal control and real-time video processing.
Onboard computing
A module of about 120 g, up to 100 TOPS, installable on the Matrice 4 Series and Matrice 400. In-flight inference for minimum-latency detection.
Dock-side on-edge AI
Industrial computing at the dock for video processing, automatic alarms and up to a 5× reduction in streaming costs.
Bring Your Own Model
Custom RFI models (rails, insulators, vegetation) deployed on-prem, never exposed to the cloud. Securely updatable remotely.
Visual Language Models
Natural-language search across historical archives: “show obstacles detected near km 124 in the last 30 days”.
From automatic flight to automatic monitoring.
Without automatic data analysis, autonomous image capture does not replace manual work. The pipeline we propose turns the video stream into actionable tickets on RFI systems, in a few seconds.
Acquisition
RGB tele/wide, radiometric thermal, georeferenced RTK telemetry.
On-edge inference
High-performance AI on board (~100 TOPS) or at the dock.
Secure stream
Encrypted video and telemetry to the remote operations centre over a private network or VPN.
Validation
A certified operator confirms, downgrades or requests further investigation.
Ticketing
Automatic work orders to RFI's CMMS/GIS with georeferenced attachments.
Omnidirectional obstacle sensing, essential in the railway environment.
The railway right-of-way is full of vertical obstacles (OCL masts, catenary, signalling, bridges). The advanced perception system of the Matrice 4 Series ensures autonomous operational safety even in BVLOS.
Omnidirectional vision
Multiple sensors for 360° coverage (horizontal and vertical).
Proximity LiDAR
Active detection of thin obstacles such as suspended wires (catenary, contact wire).
Dynamic path replanning
The drone changes its route autonomously when unexpected obstacles appear.
“Go to safe altitude”
Automatic safety procedures to avoid in-flight collisions.
Night operations
Obstacle sensing active even in low-light conditions.
Geofencing
Configurable exclusion zones: tunnel profiles, bridges, restricted areas.
A single operations centre orchestrates the entire fleet.
The orchestration software we propose is designed for critical enterprise environments. It turns stand-alone stations into an enterprise-ready network: native integration with RFI systems, BVLOS compliance, data security as a design priority.
Command & Control
- Real-time fleet status dashboard
- Launch of scheduled or on-demand missions
- Certified operator telepresence
- Automatic multi-dock deconfliction
- Manual override for anomalies
Data & AI Platform
- Structured post-mission ingestion
- Versioned, georeferenced archiving
- Batch and streaming inference
- On-edge AI detection + private cloud
- APIs to RFI's CMMS / GIS / BIM
BVLOS & Safety
- Dynamic path planning
- Configurable geofences and no-fly zones
- Integration of Detect & Avoid sensors
- SORA procedures / standard scenarios
- Audit log and full traceability
Find out how the operations centre works 24/7 today for our clients.
In-house or DaaS. Two ways to activate the network.
RFI can build and manage the network in-house, or entrust it to DroneBase with a turnkey service (DaaS). Below, the eight operational dimensions compared.
In-house management
RFI builds and manages the fleet and the operations centre in-house.
- High initial CAPEX (hardware, operations centre, training, certifications)
- Time-to-operations of 8–12 months to reach steady state
- Recruitment, training and certification of dedicated operators
- Technology updates borne by RFI (continuous refresh)
- Scalability tied to budget and internal headcount
- Operational risk borne entirely by RFI
- Centralised governance but rigidity when the scope changes
- Need to develop skills in rapidly evolving technologies
DaaS · DroneBase
Turnkey service with measurable SLAs and a predictable fee.
- Zero or minimal CAPEX, OPEX fee-based model
- Time-to-operations of 30–60 days for operational pilot lots
- Operators already ENAC-certified (Italy), with redundancy and proven processes
- Continuous technology updates included in the service
- Modular scalability by corridor or geographical area
- Operational risk transferred through contractual SLAs and KPIs
- Data ownership remains with RFI · DroneBase is only the data processor
- Scope flexibility: services can be activated / deactivated by area
| Dimension | In-house management | DaaS · DroneBase |
|---|---|---|
| Initial CAPEX | High · hardware, operations centre, training | Zero / minimal · fee-based model |
| Time-to-operations | 8–12 months to steady state | 30–60 days for operational pilot lots |
| Technology updates | Borne by RFI | Continuous, included in the service |
| Operational risk | Borne by RFI | Transferred to the supplier through SLAs and KPIs |
| Data ownership | RFI | RFI · DroneBase is the data processor |
| Scope flexibility | Low once sized | High · modular services by area |
Eight contexts where the autonomous drone creates value.
Eight concrete scenarios where autonomous stations turn personnel-intensive operations into continuously monitored workflows.
Live track
Overhead contact line
Bridges & viaducts
Tunnels
Vegetation
Yards & depots
Topography & 3D
Event response
An approach in progressive lots, validated in the field.
We start with a line pilot, collect real datasets, validate the custom AI on RFI use cases, then extend corridor by corridor up to the national roll-out. Four phases, measurable KPIs at every step.
Line pilot
1–2 monthsSelection of 1 pilot line section, 2–3 Docks + Matrice, activation of a trial operations centre, KPI definition.
Validation
2–4 monthsLive operation, collection of real datasets, training of the custom RFI AI, SLA refinement.
Strategic corridor
4–8 monthsExtension to a full corridor, integration with RFI's CMMS/GIS, operations centre scale-up.
National roll-out
8–12 monthsProgressive extension to the priority routes, stable operational governance.
An industrial partner, not a proof of concept.
Six elements that make the difference between an interesting proposal and an industrial one. Certified hardware, proven processes, commercial partnerships already in operation, direct Italian support.
Super Gold
DJI Authorized Dealer · highest channel tier
ISO 9001
Certified 2015 quality management system
DaaS
Surveillance service already in operation · SSG partnership
Academy
In-house training centre for operators and clients
R&D
Pipeline on AI, LiDAR and industrial sensors
IT
Rimini headquarters · nationwide coverage · Italian support
Frequently asked questions about railway infrastructure monitoring with drones
What can an autonomous drone inspect along a railway line?
There are nine families of activity: track and rail inspection (turnouts, switches, joints, fastenings), vegetation and right-of-way, DTM/DSM topographic survey, 3D survey and digital twin, overhead contact line including thermography, civil structures such as bridges, viaducts and tunnels, security and intrusion prevention, rapid post-event response and recognition AI. Everything is carried out from the ground, with georeferenced data that can be compared over time.
Can railway inspection with drones take place in BVLOS?
The system is designed to operate in BVLOS in compliance with ENAC (Italy). The operations centre manages dynamic path planning, configurable geofences and no-fly zones, SORA procedures or standard scenarios, audit log and full traceability. The Matrice 4TD has omnidirectional vision and proximity LiDAR, which also detects thin obstacles such as the catenary, with sensing active even in low-light conditions.
How does it integrate with existing maintenance and surveillance systems?
The operations centre exposes APIs to CMMS, GIS and BIM and provides integrations with UTM, VMS, CCTV and SAP PM via REST, WebRTC and MQTT. AI detections, validated by a certified operator, become automatic work orders with georeferenced attachments. Deployment can be on-prem or air-gapped.
What is the difference between in-house management and DaaS?
With in-house management the infrastructure manager builds and runs the fleet and the operations centre, trains and certifies operators and bears technology updates and operational risk: high initial CAPEX and 8–12 months to reach steady state. With DroneBase DaaS the service is turnkey, on an OPEX fee, with operators already ENAC-certified (Italy), contractual SLAs and KPIs and 30–60 days for operational pilot lots. Data ownership remains with the client: DroneBase is the data processor.
How long does the pilot last and how do you get to roll-out?
The roadmap has four phases: line pilot in 1–2 months (1 pilot line section, 2–3 Docks with Matrice, trial operations centre, KPI definition), validation in 2–4 months, extension to a strategic corridor in 4–8 months and national roll-out in 8–12 months. Every step has measurable KPIs.
What does the AI detect and where is the data processed?
The AI automatically classifies defects, components and obstacles and produces georeferenced output ready for ticketing. Inference takes place on board the drone, with a module of about 120 g and up to 100 TOPS, or at the dock: custom models are installed on-prem and are not exposed to the cloud. A certified operator confirms, downgrades or requests further investigation before the work order.
Why DroneBase
Since 2012, Europe’s reference point for professional UAV technology
DJI Super Gold Dealer, the only one in Europe. Headquarters, warehouse, service centre and demo area in Rimini, Italy: you see the system working before you buy it, and this is where it comes back for maintenance and updates.
- ISO 9001 certified — certified quality in sales, training and service
- Training with certified instructors — DroneBase Academy, pilot certificates and STS scenarios
- Technical and regulatory consulting on complex scenarios, ENAC and EASA
- R&D for complex projects and tailor-made integrations
- Full support, before and after the sale, with a dedicated technical contact
Let's build the first pilot lot together
Joint technical workshop, selection of the line section, Dock sizing, definition of KPIs and SLAs. A technical and financial proposal in two variants — assisted in-house and turnkey DaaS — within 30 days of kick-off. Fill in the form: one of our specialists takes charge of your request and gets back to you with technical and commercial information and, if needed, to arrange a consultation or a demo. No commitment.
