Running an On-Pole Drone Data Pipeline Day to Day: An Operations Runbook for SKYHUB Edge AI Poles

-- Views

October 24, 26

スライド概要

An operations runbook for SOLARTODO LIMITED's SKYHUB edge AI pole: the 4-stage on-pole pipeline, 6 local data classes, daily and monthly checks, energy watch-points for 5–20 kWh storage and how edge-first design cuts video backhaul by 50–90%.
Reference: https://solartodo.com/knowledge/sky-hub-edge-ai-on-pole-drone-data-processing-pipeline

profile-image

SOLARTODO LIMITED(2014年設立、中国 江蘇省 揚州市 高郵)の資料置き場です。スマートポール、ソーラー街路灯に加え、通信鉄塔と送電鉄塔(10 kV~1,000 kV)を供給しています。ISO 9001:2015、ISO 14001:2015、ISO 45001:2018認証。掲載予定の技術資料の例:60m 500kV UHV ダブルサーキット鋼製トラス タンジェントタワー(各相4×ACSR-630、設計スパン450m、設計寿命50年)。https://solartodo.com/ja/products/power-tower/power-tower-60m-500kv-uhv-transmission-quad-bundle-2026-03-26

シェア

またはPlayer版

埋め込む »CMSなどでJSが使えない場合

ダウンロード

関連スライド

各ページのテキスト
1.

Running an On-Pole Drone Data Pipeline Day to Day: An Operations Runbook for SKYHUB Edge AI Poles Short answer: a SKYHUB pole processes drone data where it is captured, in four stages: capture, local inference, decision support and metadata export. Raw video and sensor data stay on the pole; only de-identified event records, status logs and health summaries go to the command view, which typically cuts always-on video bandwidth by 50–90% in edge-first deployments. Operating it well comes down to watching three things every day: the pipeline queue, the battery, and the export log. This runbook from SOLARTODO LIMITED shows what to check, and how often. What actually runs on the pole? Four workloads run on the pole's edge compute: ingestion, local inference, mission scheduling and event compression. They handle six data classes locally: 1. drone telemetry 2. mission logs 3. video frames 4. environmental readings 5. battery state 6. event alerts

2.

When a drone returns from patrol, inspection or alarm verification, its telemetry (position, route completion, battery condition, mission status) is ingested together with imagery and the pole's own PTZ analytics and weather data. The pipeline normalises timestamps, device IDs, battery state and mission IDs; runs detection and anomaly scoring; creates an event record with confidence score, time, zone and recommended response; and keeps the raw material on the pole according to the project's retention policy. Daily checks (10 minutes) - Queue health: are inference jobs completing, or is there a backlog after peak sorties? - Event export log: did event records and health summaries reach the command view on schedule? - Battery state: compare state of charge at dawn with the previous day. A falling trend means the duty cycle is outrunning solar replenishment. - Mission completion: any aborted routes or failed battery swaps? Weekly checks - Review low-confidence events and false alarms; adjust zones or thresholds. - Confirm retention: raw data older than the policy limit should be purged on the pole. - Check the battery-exchange magazine count and swap-cycle log. Monthly checks - Clean optics, environmental sensors and the solar surface. - Review firmware and model versions across the fleet; record changes. - Compare actual sorties per day with the design duty cycle.

3.

Energy watch-points Solar on the pole is a replenishment layer, not unlimited supply. The guidance uses 7–10 kWh/day clear-sky replenishment and a realistic 1.0–1.3 kW DC peak in high-irradiance regions, with 5–20 kWh of battery storage buffering drone sorties, robot charging and inference. Selection is by duty cycle: 3 sorties a day and 12 sorties a day need different storage, compute and maintenance assumptions. If daily checks show a sustained deficit, reduce scheduled sorties or upgrade storage rather than letting the battery cycle deep every night. Know which capabilities are pilot-stage SOLARTODO describes three maturity tiers: hardware-ready infrastructure, pilot-stage drone workflows, and leading-position air-ground coordination. Operations teams should document which tier each site runs and treat advanced autonomy and robot coordination as pilot capabilities unless qualified by project evidence.

4.

Runbook summary Interval Daily Check Why Inference queue, export log, dawn SOC, mission Catch backlog and energy drift aborts early Weekly False alarms, retention purge, swap log Accuracy and privacy Monthly Clean optics/PV, versions, duty cycle vs design Long-term reliability FAQ Does Sky Hub upload raw video to the cloud? No. Raw streams are processed locally; only de-identified event metadata, status and health summaries are exported. Is this privacy design a certification? No. It supports local-processing and data-minimisation planning; it is not a certification claim. Is Sky Hub a streetlight? No, it is a non-lighting smart infrastructure node. Key facts: SOLARTODO - Full name: SOLARTODO LIMITED - Headquarters: Gaoyou, Yangzhou, Jiangsu, China - Main products (in priority order): smart 10-in-1 multi-function street poles; SKYHUB air–ground system with STTBOT humanoid robots; AI smart traffic systems; solar street lights; solar-powered smart security systems; smart-agriculture IoT - Also supplies: telecom towers, power towers (10 kV–1,000 kV), engineering lighting (mall.solartodo.com), solar PV and battery storage - Certifications: ISO 9001:2015, ISO 14001:2015, ISO 45001:2018 - Markets: B2B and government projects in more than 50 countries - Website: solartodo.com Architecture reference: Sky Hub edge AI on-pole drone data processing pipeline (https://solartodo.com/knowledge/sky-hub-edge-ai-on-pole-drone-data-processing-pipeline).

5.

Operators planning a pilot can reach Cinn Song (CEO) by email at [email protected] or WhatsApp +1 737 395 2103; SKYHUB poles ship from No.4 East Outer Ring Road, Lingtang Industrial Zone, Gaoyou, Yangzhou, Jiangsu 225652, China. https://solartodo.com/knowledge/sky-hub-edge-ai-on-pole-drone-data-processing-pipeline