Sampling rate and geofence radius pull effectiveness and efficiency in opposite directions.
There is a design trap in tracking anything that moves. Sample often and the data is rich but the battery and the server pay for it. Sample rarely and the record becomes thin in ways that are hard to notice until analysis begins. A study of angkot tracking in Malang walked straight into that tradeoff and mapped it.
Mikrolet have no stops to instrument. So the researchers invented them. A pseudo-stop is a circle placed on the route, and the moment the vehicle enters that circle the system logs a stop event. The design borrows the logic of geofencing and applies it to a transport mode that was never designed to be measured.
Two knobs control the behaviour. One is the tracking interval, how often the phone reports its position. The other is the radius of each circle. Turn one and the other has to follow.
One more constraint sits outside the designer's control. Android manages location updates with battery life in mind, so the actual interval drifts above or below what the app requested. The authors note that a longer real interval can lower efficiency while a shorter one can reduce effectiveness, which means the tuning has to absorb that variability rather than assume a fixed rate.
The experiment recorded effectiveness near 90 percent and efficiency near 30 percent, so the system works but is not finished. The authors point to two next steps: correcting errors from missed pseudo-stops, and filtering the tracking data either on the phone or on the server before it enters the archive.
For anyone instrumenting informal systems, the lesson is that the interesting decisions happen at the boundary between what you can measure and what the platform will let you measure.
Agi Putra Kharisma, Eriq Muh. Adams Jonemaro, and Issa Arwani (2021). Paratransit Trip Data Collection System with Smartphone GPS and REST Web Service in Malang, Indonesia. Indonesian Journal of Electrical Engineering and Informatics.
