Infrastructure inspection is where a drone earns its keep by going somewhere awkward. Unfortunately the same features that make a site worth inspecting, such as steel, concrete, height and enclosure, are the ones that interfere with satellite navigation. GPS needs a clear view of several satellites spread across the sky and a direct signal path from each. Inspection sites break that in three ways. Structures block the sky, leaving too few satellites or a poor geometry. Large flat surfaces reflect signals, so the receiver sees a delayed copy alongside the direct one, which is called multipath and produces position jumps rather than an outright loss. And electrical infrastructure can add radio noise near the GPS bands. The result is rarely a clean loss of signal; more often it is a fix that is present but wrong, which is worse. This article looks at the common GPS-hostile inspection environments and how a mission can move between segments where GPS is fine and segments where it is not.
Bridges and the underside problem
The most valuable bridge inspection images are of bearings, girders, pier caps and the underside of the deck. Flying there puts a deck of concrete or steel directly overhead and often a river or gorge below. The sky view collapses, multipath from the deck and piers is severe, and a compass may swing near steel. Many operators handle this today by flying manually with GPS modes disabled, which is demanding work and hard to repeat consistently. A navigation source that takes its reference from the ground and the surrounding terrain rather than the sky changes that picture, though under a wide deck even the ground view is constrained and the inertial bridge has to carry more of the load.
Urban canyons
Between tall buildings the sky is a narrow strip, satellites appear and disappear as the aircraft moves, and glass facades are efficient reflectors. Position estimates in these conditions can wander, then snap back when the geometry improves. For facade, roof and utilities inspection in dense areas, the practical concern is not just accuracy but predictability: an aircraft that suddenly believes it has moved sideways can react in ways the pilot does not expect. A terrain and imagery reference sees roads, rooftops and block layouts that are highly distinctive, which makes cities, perhaps surprisingly, good ground for visual matching even while they are poor sky for GPS.
Linear assets: pipelines and power lines
Pipelines and transmission lines are inspected over long distances and the challenges are different. Most of the corridor may have a fine GPS view, but the route will pass through gullies, forest, cuttings, substations and river crossings where coverage drops. A fixed-wing aircraft is the natural platform for these missions because of its range, and it will cross these problem segments at speed. What matters is that the navigation solution keeps track through each one and comes out the other side still aligned with the corridor, so that imagery is correctly geolocated and the aircraft holds the required offset from the asset.
Linear corridors also tend to have strong terrain and visual features along them: the cleared easement itself, access tracks, towers, valve stations and creek crossings. That helps a terrain-referenced system considerably.
Transitioning between GPS-available and GPS-denied segments
The interesting part of a mixed-coverage mission is the handover, in both directions. Several things need to be true for it to work well:
- The secondary source must already be running and converged before GPS degrades, not started in a hurry when it fails. Warm and continuous is the goal.
- The autopilot’s fusion filter should weight sources by their reported confidence, so the transition is a gradual shift in trust rather than a hard switch that jolts the position estimate.
- Degraded GPS must be detected as degraded. A fix with poor geometry or multipath that still reports a healthy status is the classic trap, and cross-checking it against an independent absolute source is one of the few ways to catch it.
- When GPS returns, the two sources should agree closely before it is trusted again, so a spoofed or multipath-corrupted fix cannot drag the estimate away on re-acquisition.
- The remote pilot should see which source is primary at any moment and be able to force a choice manually.
Good practice is to map the expected GPS-hostile segments during planning, using the site’s geometry and any previous flight logs, and to treat each one as a phase with its own entry and exit conditions. Approach the segment on a clean fix, enter with the terrain source confident, keep the flight path within ground the terrain database covers at adequate resolution, and exit onto a section with open sky where both sources can be checked against each other before the next phase. Where a structure obscures the ground as well as the sky, such as directly beneath a wide bridge deck, plan the segment to be short and accept that the inertial estimate carries it.
Where TerrainSLAM fits
TerrainSLAM provides an absolute position derived from the ground rather than the sky, computed onboard and delivered to the autopilot as an independent source that can be cross-checked against GPS and take over when GPS degrades. Its imaging pipeline is designed to cope with haze, shimmer and changing light across the altitude range from low structure passes to higher corridor surveys, and its autopilot-agnostic interface is intended to make the source-switching logic described above possible on existing platforms.