A Week on the Move: What a Mobile LoRa Receiver Heard Across Taipei

We took a Meshtastic receiver on the road for seven days. Here’s what 1,643 packets, 226 nodes, and one very well-traveled node named ure5 taught us about our local mesh network. This is part of the Self-Organized LoRa Emergency Communication Drill series during Urban Resilience Exercise 2026 in Taiwan.

Most of the time, when we log Meshtastic traffic, we do it from a fixed spot — an antenna on a rooftop, quietly listening to whatever drifts by. This time we did something different: we let the receiver ride along, moving through Greater Taipei (and a couple of longer trips beyond it) over seven days, from August 7 to August 13, 2026.

The result is a much messier, much more human dataset than usual — and honestly, a lot more fun to dig through.

The numbers at a glance

Metric	Value
Total packets captured	1,643
Unique sender nodes	226
Listening period	7 days
Average SNR	-3.1 dB
Median distance	8.7 km
Farthest reception	105.6 km
Text / chat messages	781

For comparison, our fixed station (ure1) picked up 115 unique nodes over roughly 12 hours. Seven days of mobile listening nearly doubled the node count — moving around really does get you a bigger sample of the network.

Activity built up day by day

Both the packet count and the number of unique nodes climbed steadily through the week, peaking hard on the last day — August 13, with 483 packets and 128 unique nodes in a single day. Whether that’s because the receiver rolled into a signal-rich part of town or because that day just had more people on the air, it’s a good reminder that a single day’s snapshot can badly undersell (or oversell) how “busy” a mesh network really is.

It’s not just position pings — people are actually talking

This was the biggest surprise in the dataset: text and chat messages (781) actually outnumbered standard POSITION_APP broadcasts (593). Digging into those messages turned up:

• Bot traffic — TPE-BOT, a relay bot based in Wanhua, echoing hop counts and SNR for other nodes; and MeshAI, a rooftop-hosted assistant that answers @mesh queries about weather, PM2.5, earthquakes, and node temperature.
• MQTT bridge chatter — a node nicknamed Fanzaiyuan (板橋浮洲) relaying pings, distances, and hop counts from nodes connected via MQTT.
• Plain old testing and small talk — @ab, test, Hi, 安安 (a casual Taiwanese “hey”), and a thumbs-up emoji here and there.

In other words, this isn’t just a network of devices quietly reporting GPS coordinates — it’s a living community using LoRa mesh for genuine, if lightweight, communication.

Who’s talking the most

The node ure26-field-8 topped the activity charts with 128 packets, followed closely by 🐱AK-HELTEC-V4 and MeshAI. Several of the most active nodes are bot or bridge nodes rather than simple end-devices — a sign that this part of the mesh has some real infrastructure behind it, not just individual hobbyist radios.

Signal quality: more variable on the move

Average SNR came in at -3.1 dB, across a noticeably wider range than our fixed station saw (-16.5 dB to +15.5 dB). That’s exactly what you’d expect from a moving receiver: sometimes you’re right next to a strong node, sometimes you’re at the ragged edge of coverage. It’s a nice illustration of why mobile surveys are useful for mapping real-world coverage gaps — a fixed station only ever tells you about its own neighborhood.

Where the receiver actually went

This is the chart unique to a mobile listening session — the path the receiver itself traveled, colored from dark purple (start) to yellow (end) to show the order of movement. Most of the time was spent in and around Banqiao and Zhonghe in New Taipei (see the zoomed inset), but the log also shows a couple of longer excursions out toward Taoyuan and Hsinchu, reaching as far as roughly 121.03°E, 24.81°N. Coverage kept getting logged even during those longer trips — a nice stress test of how far the mesh actually reaches.

Where the other nodes are

Plotting the senders themselves shows the network is heavily concentrated around Banqiao, Zhonghe, and Xinzhuang — right where the receiver spent most of its time — with a scattering of nodes further out toward Hsinchu and Yilan. The dashed line is the receiver’s own path for reference, so you can see how closely (or not) the listening route tracked the densest part of the network.

What this tells us

A few takeaways from this round of listening:

1. Mobile logging finds more of the network. Nearly 2x the unique nodes compared to a fixed station, in a comparable or shorter listening window.
2. This mesh is genuinely social. More than half the non-APP payloads were people (or bots) actually talking, not just machines reporting position.
3. Coverage is real but uneven. SNR swings widely depending on where you are relative to the node cluster — useful information if you’re planning where a fixed relay station might help most.
4. Bots and bridges matter. Infrastructure nodes like MeshAI and the MQTT relay are pulling real weight in keeping the network useful day to day.

Next step: we’d like to line this mobile dataset up against fixed-station logs (ure1, ure2, ure4, ure7) from the same week, and see whether the activity spike on August 13 lines up with a specific event or drill.

---

This report is part of an ongoing series of self-organized LoRa emergency communication drills.

Questions or want to compare notes? Reach out at tak.ywtcu@simplelogin.com
More background: ure26.takke.me

— TAKKE.me