Haiduan is spending NT$7.5 million to geocode streetlights and farm roads: the real asset is a maintainable public-service base map
Original Chinese title: 海端用750萬元把路燈、農路做成可定位資料：55原鄉需要的不是一張地圖，而是可維護的公共服務底圖
Haiduan Township Office posted a NT$7.5 million joint procurement on September 7 for a streetlight inventory, asset tags and management system, together with a farm-road survey and GPS mapping. The policy value is not merely putting infrastructure on a map, but creating location data that can support maintenance, disaster reporting, agricultural access and budget planning.
Yuan Media AI Editorial Desk
The Yuan Media AI Editorial Desk covers public policy, Indigenous communities, cultural continuity, agricultural resilience, digital governance, and AI-enabled public services across Taiwan's 55 Indigenous areas.
On September 7, 2026, Haiduan Township Office in Taitung posted a joint service procurement with a total budget of NT$7.5 million, tender number HSW1150831. The procurement is divided into two parts: NT$3.5 million for a streetlight basic-data inventory, physical asset tagging and an information-management system, and NT$4 million for a farm-road survey and GPS mapping. The published tender information lists a bid deadline of 5:00 p.m. on September 17 and a scheduled opening on September 18. The streetlight component is to be completed within 240 calendar days from the day after award, while the farm-road component has a 250-day performance period.
It would be easy to read this as an ordinary surveying or software procurement. Its larger relevance to Taiwan's 55 Indigenous areas is the problem of the public-service base map. Mountain roads, farm roads and streetlights are spread across settlements, slopes, river valleys and long transport corridors. If asset information remains fragmented across paper files, individual staff computers and one-off construction records, each maintenance request begins by rediscovering location. After a disaster, the township may also struggle to answer a more basic operational question: exactly which road segment, light or agricultural access point has been affected?
The streetlight component matters because an inventory plus a physical tag can give each real-world fixture a stable identity. A resident should not have to describe a failed light as "the third bend above the settlement" if an asset number or mapped point can identify it. On the administrative side, maintenance history, fixture type, failure frequency and last-update date can be attached to the same asset. If the data is maintained, that foundation can support repair dispatch, night-time safety, energy-efficiency replacement and annual budget prioritization. If the survey is treated as a one-time deliverable, however, the map will gradually become another outdated dataset.
The farm-road GPS component has a direct connection to agricultural resilience. In a geographically large mountain township, farm roads are not only routes for moving produce. During heavy rain, slope failure or road disruption, they can also serve as essential background for field reporting, damage assessment and the interpretation of possible alternative access. A common coordinate system can allow agriculture, public works, disaster management and local service teams to refer to the same segment. But a GPS line must never be mistaken for real-time road status. Survey accuracy, update time, current passability and field verification need to remain separate fields.
This procurement also highlights a more difficult issue: data governance matters more than the map interface. First, the township needs stable identifiers and a small set of required fields, so the same streetlight or road segment does not acquire different names in different systems. Second, maintenance responsibility must be explicit. When a light is replaced, a road is rerouted, construction is accepted or post-disaster repairs are completed, somebody must update the corresponding record. Third, public-service information should be separated from internal operational data. Residents may need asset identification, reporting functions and service progress, but that does not mean every precise infrastructure location, maintenance attribute or system privilege should be openly exposed.
The public procurement record also marks this tender in fields related to sensitivity, national security, including cybersecurity concerns, and national-security involvement. That makes account control, permissions, backups, exports, maintenance access and vendor handover especially important acceptance issues. A system should not be considered complete simply because the map opens in a browser. The township should be able to restore its data, trace changes, revoke access, export core records and continue operations if a contractor changes.
AI should not be forced into the headline of the underlying procurement. The actual tender is about field inventory, GPS surveying, geolocation and information management, not an AI system. AI may become useful after the base data is stable: for example, matching a resident's repair message to an existing asset, organizing post-disaster photographs by road segment, identifying repeated failures, or generating a queue of cases that require human review. Any automated output should remain anchored to verified asset identifiers and timestamps. A generative model must not be allowed to invent a road or streetlight simply because a location description is incomplete.
For the 55 Indigenous areas, this is a more important digital-governance direction than building another promotional app. Long-term local digital public service requires a maintainable location layer in which roads, farm roads, streetlights, evacuation points and essential public facilities follow consistent location and identification rules. Separate service systems can then use the shared base data according to their own permissions. This reduces the risk that disaster response, agricultural access, repair services and local construction each build a separate data island.
The policy value of Haiduan's NT$7.5 million procurement therefore lies less in the headline budget than in what happens after the first survey. Three questions deserve continued scrutiny: who updates the data after acceptance, how residents and staff use the same traceable identifiers, and whether the system remains usable under weak connectivity, disaster conditions and cybersecurity constraints. If those questions are answered clearly, the project can become an enduring public-service capability rather than a one-time digital installation.
Official and public procurement sources
AI use and content-safety disclosure
This article is based on Haiduan Township Office's official website and public procurement data. AI is used only to organize data-governance and public-service analysis; it does not replace procurement documents, survey results, acceptance testing or administrative decisions.