<![CDATA[Purpose – Reliable passenger data are important for occupancy monitoring, operational supervision, and service reporting in public bus transportation. This study designs and conducts a laboratory evaluation of a microcontroller-based automatic passenger-counting prototype for DAMRI bus capacity monitoring, with particular attention to context-dependent correction of non-revenue passenger movements. Design/methods/approach – The prototype integrates an ESP32 microcontroller, two vertically positioned HC-SR04 ultrasonic sensors, an LCD, an overload buzzer, an event-aware dual-mode intervention button, and Telegram Bot connectivity. Passenger direction is identified from the sensor activation sequence. Following a detected event, a short button press reverses the most recent boarding or alighting event when the detected individual should not be included in the official passenger count, whereas a long press transmits the operational report and resets the session counters. The revised evaluation included positive-event sensor-detection trials, boarding and alighting classification, complete non-revenue boarding–alighting sequences, zero-occupancy safeguard testing, capacity-warning tests, intervention-button operation, and Telegram communication. Findings – In the original controlled positive-event trials, the sensor-detection success rate was 94% (47/50; 95% Wilson CI: 83.8–97.9%). Boarding and alighting classification achieved 92% (46/50; 95% CI: 81.2–96.8%) and 94% (47/50; 95% CI: 83.8–97.9%), respectively. The revised intervention mechanism correctly handled 10/10 complete non-revenue movement scenarios, while the zero-occupancy safeguard succeeded in 10/10 tests. The capacity-warning mechanism performed correctly in 10/10 threshold scenarios. Telegram monitoring achieved 10/10 successful authorized responses, with a mean response latency of 1.04 s. Research implications/limitations – The prototype demonstrates the feasibility of integrating directional counting, context-aware manual correction, local capacity alerts, and remote reporting within a low-cost embedded platform. However, testing remained laboratory-based and did not reproduce the full variability of actual bus operation. Simultaneous multi-passenger movement was not evaluated. Originality/value – The contribution does not lie in the established two-sensor directional-counting principle. Instead, it lies in integrating this sensing method with an event-aware bidirectional intervention and reporting workflow designed to correct context-dependent non-revenue passenger movements in Indonesian bus operations.]]>
