Ten walkers, two search tasks and one chart of frustrations produced a bus finder that cut travel time by roughly two thirds.

A pedestrian hunting for a bus stop is not driving. There is no lane to follow and no turn coming up. The person stands on a pavement, turns in a circle, and tries to line up a phone screen with the street in front of them. That mismatch is where location based AR usually stumbles.

The authors ran a field study with ten university students aged 20 to 28. All used Android phones, and none knew the test area. Each participant completed two tasks. In blind search the app named a random destination about 300 metres away. In directed search the participant picked a place from the map. Every step was recorded.
Those records became a user journey map. Along the top sits the sequence of steps a user takes. Down the side sit five rows: actions, questions, happy moments, pain points and opportunities. The last two rows are where the software requirements come from.

The team built an AR app for the bus network of Zhongli District in Taiwan and rebuilt the presentation around one rule: do not show everything at once. Only the next waypoint and the final target appear. When two markers line up, the farther one is hidden. A marker changes shape once the user is within 20 metres of the destination.

Ten participants then repeated both tasks on the prototype and on a standard location based AR app. Both scored 100 percent success, so accuracy was never the issue. Time was. Blind search fell from 14.9 minutes to 5.4, and directed search from 7.86 to 5.12. Questionnaire scores for satisfaction and learnability climbed by roughly 15 to 25 points.
One complaint survived the redesign. Several users reported drift, where a marker floats slightly away from the real object. That is a sensor and GPS limit rather than an interface one, and the authors note that more precise coordinates would be needed to fix it.
Brata, K. C., & Liang, D. (2019). An effective approach to develop location-based augmented reality information support. International Journal of Electrical and Computer Engineering.