Transport connectivity is an engineering problem of velocity. A router that performs beautifully on a bench has never had to hold a session through a mast handover at 100 km/h, from inside a steel box, on a power supply that sags every time the doors cycle. Fleet networks fail in ways static networks never see, and they get designed accordingly.
What speed does to a network
The core difficulty is not signal strength, it is handover stability. A static router keeps a settled relationship with one mast; a moving vehicle renegotiates that relationship constantly, and when the handover is slower than the movement, sessions die: VPNs drop, CCTV freezes, ticketing transactions fail mid-payment. At speed, rapid attenuation can push weaker modems into a ping-pong between towers without ever settling on either.
The vehicle itself is the second adversary. A bus or rail chassis is a Faraday cage, so internal antennas are fighting the bodywork before they fight the RF environment; external roof-mounted MIMO arrays are the fix, not an upgrade. And the electrical environment is hostile in ways datasheets gloss over: ignition transients and regenerative braking produce voltage swings that reboot consumer-grade kit mid-route. We specify hardware with wide voltage input, ignition sensing for graceful shutdown that protects the battery overnight, vibration-rated M12 or screw-on SMA connectors, and certifications appropriate to the platform, E-Mark for road vehicles, EN 50155 for rail.
Multi-carrier bonding on the move
No single UK operator has the best network everywhere along a route; the carrier that owns the city centre often disappears in the rural stretch or the cutting. Multi-modem hardware with SIMs across operators, bonded by SpeedFusion into one logical tunnel, means the vehicle continuously draws from whichever paths are healthy at that moment on the map. The tunnel persists while the physical links churn underneath it, which is the property that keeps CCTV streaming, telematics reporting and ticketing transacting through every handover.
For the traffic that cannot tolerate a gap, WAN Smoothing duplicates packets across links so a momentary drop during a handover never reaches the application, and Hot Failover holds standby paths live rather than cold. Bonding also solves the capacity problem that passenger expectations create: combining multiple 4G and 5G links yields the headroom for hundreds of concurrent Wi-Fi users without starving the systems the operator actually runs the service on.
Passengers and operations must never compete
The most common design failure I see in fleet networks is treating the vehicle as one network. It is two, with opposite requirements. Operational traffic, CCTV, telematics, driver comms, payments, is low-volume and critical; passenger Wi-Fi is high-volume and expendable. We segment them with VLANs and firewall policy so they never share fate: bandwidth priority guarantees the operational side whatever it needs first, passenger traffic gets what remains, and a security incident on the public side has no route to the systems running the vehicle. Per-network data policies also keep SIM costs sane; passenger video streaming should never be the reason an operational data pool runs dry.
Running a hundred vehicles from one screen
At fleet scale, the estate is the product. InControl2 gives an operations team per-vehicle link health, GPS position, and per-SIM consumption in one view, with configuration templates so vehicle 101 enrols identically to vehicle 1, and firmware staged in waves across the fleet during depot hours. The GPS-correlated history earns its keep quickly: it maps exactly where along each route the network struggles, so recurring blackspots become engineering inputs, adjusting carrier mix or antenna configuration, rather than recurring complaints.
Scoping a fleet deployment
We scope from the routes and the applications: what must survive at speed, what the realistic coverage looks like along the actual corridors the fleet runs, what the passenger load demands, and what the vehicle platform imposes electrically and physically. Then hardware, carrier mix and tunnel design follow. We build the configuration templates, prove the design on pilot vehicles, and train the operator's team to run the estate, because a fleet network handed over without transfer of understanding is a support contract in disguise.
The short version
Bond across carriers so no operator's coverage map defines your service. Put the antennas outside the steel. Specify for the electrical reality of a vehicle, not a bench. Keep passengers and operations on separate networks with the priority written down. And run the fleet from one screen with templates and staged rollouts. If your fleet's connectivity is still a per-vehicle adventure, get in touch for a scoping conversation.
Frequently asked questions
Why do connections drop at speed even in good coverage areas?
Handover, not coverage. Moving between masts forces constant renegotiation, and single-link setups break sessions when a handover stumbles. A bonded multi-carrier tunnel rides through it because the session does not live on any single link.
Can passenger Wi-Fi and operational systems really share one router?
Yes, when properly segmented: separate VLANs, firewall isolation, and bandwidth priority that guarantees operational traffic first. The failure cases you hear about come from fleets that skipped the segmentation.
What hardware certifications matter for vehicles?
E-Mark for road platforms, EN 50155 for rail, plus practical must-haves the certificates imply: wide voltage input, ignition sensing, vibration-rated connectors and antennas mounted outside the chassis.
How do you control SIM costs across a large fleet?
Per-network data policies, passenger bandwidth caps, per-SIM consumption visibility in InControl2, and alerts on anomalous usage, so an unexpected consumer is caught in days rather than discovered on an invoice.
Do tunnels and cuttings defeat this architecture?
They defeat individual links, which is the point of not depending on any. Carrier diversity narrows the blackspots, GPS-correlated history maps what remains, and buffering strategies for telemetry cover the genuinely dark sections.