// INSIGHTS
AR Wayfinding Across a Large Venue: Anchors and Fallbacks
6 min read
AR wayfinding demos beautifully and deploys badly. In a controlled corridor, an arrow floats on the floor and leads you to a door. In a 40,000 square metre expo hall with 18,000 people, glass walls, repeating truss and saturated WiFi, the same system loses its anchor in 30 metres. The technology is real and useful. It just has to be built as a navigation system with an AR layer, rather than an AR experience that happens to navigate.
The three anchoring strategies
Pure visual-inertial odometry uses the phone camera and motion sensors to track movement from a known start. It requires no infrastructure and it drifts. Expect accumulated error in the range of three to five percent of distance travelled in a feature-poor environment, which means an arrow that is two metres off after a 50 metre walk and useless after 150. It is adequate for short hops inside one hall, not for cross-venue routing.
Image-target anchoring places printed markers at known coordinates. The phone sees a marker, resets its position to zero error, and drifts again until the next one. This is unglamorous and it works. A marker every 25 to 40 metres, plus one at every decision point and lift lobby, keeps the error bounded. Markers can be brand-designed floor decals or wall plates rather than visible QR squares, and they double as static wayfinding for the majority of visitors who will never open the AR at all.
A visual positioning system pre-scans the venue into a point cloud and relocalises the phone against it. It gives the best result when it works, typically relocalising within one to two metres. It also requires access to the venue before the build is complete, which at an expo means scanning an empty hall that will look completely different once 300 stands are erected. Scanning after build-up is possible but leaves a very short window. For most Indian and UAE expo timelines, a hybrid of image targets with VIO between them is the reliable choice.
What to budget for
Build the system in layers
A 2D map that works alone
The base layer is a fast mobile map with search, floor switching and a route line. It must be complete and useful with the camera closed, because most visitors will use only this.
AR bursts at decision points
Do not ask anyone to hold a phone up for 200 metres. Trigger the camera view only at junctions, lift lobbies and hall entrances where the choice is genuinely ambiguous.
Web AR, not a native app
Browser-based AR removes the install step entirely. It costs some tracking quality and some older-device support, and it is almost always worth it because install rates at events are low.
Offline route data
Cache the full map, route graph and marker registry on first load. Venue WiFi will fail at peak, and a wayfinding tool that stops working when the hall is busiest is worse than a printed map.
Deployment sequence
Build the route graph from the floor plan
Convert the venue CAD into a node and edge graph with real distances, accessibility flags and floor transitions. Everything else depends on this being accurate to the metre.
Place and survey the anchors
Fix marker positions on the plan, install them during build-up, then walk and verify each one with the live app. A marker installed 1.5 metres from its recorded position poisons every route through it.
Walk-test every primary route
Test at least the twenty highest-traffic routes end to end, on both an iOS and an Android device, once the stands are up and the lighting is at show state.
Publish the fallback prominently
Every AR entry point should also show the plain map and the nearest static sign. Give the visitor an exit from the technology at all times.
What usually goes wrong
Adoption is the first and largest failure. Attendees at a trade expo are there for meetings, not for software, and a system that requires an install will be used by a small fraction of the hall. Web AR reached through a QR on the badge, the signage and the entrance arch performs far better, but even then, plan on a minority of visitors ever opening the camera view. That is not a reason to skip it. It is a reason to ensure the value is delivered by the map layer and the AR is an assist at the hard junctions, so the project is not dependent on adoption to justify itself.
The physical environment fights you in specific, predictable ways. Glass frontages and mirrored mall interiors confuse feature tracking. Repeating structural elements, identical stand shells and long featureless corridors cause the tracker to relocalise onto the wrong place, which is more dangerous than drifting, because the arrow becomes confidently wrong. Crowd density blocks the camera view of floor markers entirely at peak, so wall-mounted anchors at eye height outperform floor decals in busy halls. Network saturation is the other constant. Venue WiFi at an Indian expo peak hour and cellular inside a mall basement are both unreliable, which is why route data must be cached. Finally, battery. Live AR consumes a phone quickly, and a visitor whose phone dies is a visitor who cannot access their badge, tickets or contacts. Say so in the interface, keep sessions short, and put charging near the main junctions.
Common questions
Does this need beacons or extra hardware?
Usually not. Printed image anchors plus phone sensors cover most venues at a fraction of the cost of a beacon deployment. Beacons help in very large multi-level sites where you need coarse floor-level positioning before the camera opens.
How accurate will the arrow be?
Within one to two metres immediately after passing an anchor, degrading with distance from it. With anchors every 30 metres the error stays inside a few metres throughout a route, which is enough to point at the correct aisle and stand.
How long does deployment take on site?
The software is ready before load-in. On site you need roughly one day to install and survey anchors and walk-test routes in a large hall, and that day has to fall after stand build-up, which is the tightest constraint in the schedule.
Map first, AR second
Send us the venue plan and we will scope the route graph, anchor count and on-site survey window.
