AGV Roaming and Addressing on a Factory WLAN
A washing-machine plant runs 60 laser-guided AGVs along 4 m aisles between production lines. Each AGV sends 20 Hz telemetry and needs 10 Hz motion commands; the control loop tolerates 20 ms end-to-end latency and under 0.1% packet loss. The site has 420 Wi-Fi 6 access points built for office laptops and handheld scanners, all broadcasting one SSID on 2.4 and 5 GHz with 40 MHz channels. AGVs drop off the WLAN about 30 times per shift, almost always at aisle intersections, and each drop stops production for 4-8 seconds. Every device is in 10.20.0.0/24 and the DHCP pool is exhausted; 900 vibration sensors and 200 new handhelds are planned. The plant is also evaluating private 5G for the AGVs.
1. Diagnose why the AGVs disconnect at intersections and list the WLAN changes you would make.
Read the worked answer
Disconnects at intersections are the classic signature of a client choosing between two access points with nearly equal signal and re-associating slowly, not a coverage hole. Contributing factors here:
- One SSID shared with laptops and scanners means the AGV competes for airtime and inherits band steering and client limits tuned for office use.
- 2.4 GHz is used, adding interference and worse roaming behaviour.
- 40 MHz channels halve the number of non-overlapping channels, so neighbouring APs overlap and co-channel interference rises.
Changes:
- Dedicate a separate SSID and VLAN to AGVs, 5 GHz only, 20 MHz channels, with a clean non-overlapping channel plan.
- Enable 802.11k, 802.11v and 802.11r fast transition, and disable band steering for that SSID.
- Lower AP transmit power and set a minimum RSSI around -70 dBm so cells have crisp boundaries and clients move before the signal collapses.
- Apply WMM/QoS priority to AGV traffic and disable power save on the AGVs.
Then measure handover time with a roaming test, not by waiting for the next production stop.
2. Design an addressing and segmentation plan for the current and planned devices.
Read the worked answer
The current /24 gives 254 usable addresses against at least 60 AGVs, 420 APs and controllers, 500 handhelds, PLCs and HMIs, plus 900 sensors: well over 1,000 devices, so exhaustion is structural, not a DHCP bug.
Proposed plan:
- AGV segment: 10.20.10.0/26 is too tight at 62 hosts, so use /25 for 126 addresses; leave room because the fleet grows.
- Vibration sensors: 10.20.20.0/22 for 1,022 addresses, with predictable allocation by line and machine so a sensor ID maps to a location.
- Handhelds and voice: 10.20.30.0/23.
- OT/PLC segment: 10.20.40.0/24 with static addressing and a firewall between it and IT. AGVs should not sit in the PLC VLAN.
- Management: separate VLAN for APs, controllers and gateways.
Use DHCP reservations for AGVs and APs, static addresses for PLCs, and one DHCP scope per VLAN. Document the plan and reserve a growth prefix; do not simply widen the old /24 to a /16, which would create one flat broadcast domain and no security boundary.
3. Would private 5G be a better fit than fixing Wi-Fi? Argue both sides and recommend a path.
Read the worked answer
Private 5G merits evaluation but should not be the first move.
For 5G: licensed or shared spectrum gives deterministic access and no interference from office devices; 5G URLLC targets millisecond latency; handover is network-controlled rather than client-driven, which suits moving AGVs; one network can carry AGVs, sensors and video with network slicing.
Against: cost and time. 5G core, radio units and spectrum licensing are heavy for a single plant, and 60 AGVs is a small fleet with a proven alternative. AGV vendors often have certified Wi-Fi onboard; a 5G gateway adds a new failure point and a new integration project.
Recommended path: fix the Wi-Fi first with a dedicated AGV SSID, fast roaming, 20 MHz channels and QoS, and set a measurable acceptance target of under 20 ms latency and under 0.1% loss over two weeks. In parallel, run a 5G pilot in one production hall. Choose 5G if the control loop is later tightened, if the fleet grows, or if the hall cannot be made RF-clean because of welding or other interference.