IoT Railway Platform Crowd Alert

Festival seasons and evening rush hours push Indian railway platforms past comfortable — and safe — densities, yet crowd control still depends on staff eyeballing the platform. A cheap, always-on density estimate per platform zone would let staff act before a zone chokes: open another exit gate, hold passengers at the footbridge, or dispatch RPF to the dense end. This project builds exactly that as a student-scale system: HC-SR04 ultrasonic sensors mounted at intervals along a platform (or a lab-scale platform model) divide it into three zones; an ESP32 polls each sensor, smooths the readings and maps them to a 0–100 density index per zone. Crossing the warning level (60%) updates the dashboard; crossing the alert level (80%) drives a red LED-matrix sign with a scrolling message, sounds a buzzer and fires a push alert to the staff channel over Wi-Fi/MQTT. The dashboard — the software half of this hybrid build — shows per-zone gauges, a rolling 60-minute density chart, the current signage message and an escalation history. Because ultrasonic echoes measure presence in a zone rather than counting heads, the report is explicit: the system reports density bands, and the on-site calibration procedure (empty / normal / rush readings) is what makes the thresholds meaningful.

1. Each HC-SR04 pair measures distance to the nearest bodies in its zone; shorter, noisier echoes indicate a denser crowd.
2. The ESP32 samples every few seconds, applies a moving average, and converts the smoothed distance into a per-zone density index (0–100).
3. The index is compared against calibrated thresholds: below 60% is normal, 60–80% is warning, above 80% is alert.
4. On warning, the dashboard gauge changes state; on alert, the LED matrix switches to the alert message, the buzzer sounds and a push alert fires.
5. Every reading, threshold crossing and signage change is published over MQTT to the dashboard's gauges, chart and event log.
6. Staff acknowledge or escalate from the alert console; the escalation history records what happened, when, and in which zone.
7. If Wi-Fi drops, the node keeps sensing and driving the sign locally; telemetry backfills when the link returns.

Tech stack:

• ESP32 DevKit (Arduino-core firmware, Wi-Fi + MQTT)
• HC-SR04 ultrasonic sensors (6x, two per zone)
• 32x8 red LED matrix display module + buzzer
• MQTT broker + web dashboard (HTML/CSS/JS, MQTT over WebSockets)
• 5 V USB / regulated supply for the node

• [Zone-wise ultrasonic density sensing] Three platform zones, two HC-SR04 sensors per zone, polled and smoothed by the ESP32 into a 0–100 density index.
• [ESP32 alert firmware] Wi-Fi/MQTT telemetry, threshold state machine (normal / warning / alert), and local decision-making that keeps working if the dashboard link drops.
• [LED-matrix crowd signage] 32x8 red LED matrix scrolling context messages ("CROWDED — USE EXIT GATE") with ambient-light brightness control.
• [Buzzer + staff push alerts] Local buzzer on alert level plus push notification to the station-staff channel with zone and density.
• [Live crowd dashboard] Per-zone occupancy gauges, overall crowd status pill, 60-minute density chart, current signage preview and timestamped alert log.
• [On-site threshold calibration] Empty/normal/rush baseline procedure that sets the warning (60%) and alert (80%) levels for the actual installation geometry.
• [Alert rules console] Editable thresholds, message queue for the LED sign, and an escalation history view for post-incident review.

• Complete ESP32 firmware (sensing, smoothing, threshold state machine, MQTT, signage driver)
• Sensor node + LED-matrix signage hardware, wired and tested
• Crowd dashboard web app (gauges, chart, alert console, escalation history)
• Wiring diagram and sensor-mounting guide for the platform/model geometry
• Threshold calibration guide (empty/normal/rush baseline procedure)
• Project report PDF (background, sensing method, firmware design, honest accuracy discussion)
• PPT presentation for final review
• Viva Q&A preparation document (ultrasonics, ESP32, MQTT, threshold design, calibration)
• Setup guide (broker setup, Wi-Fi provisioning, first calibration run)

• Ultrasonic sensors estimate density bands — they do not count people. Exact headcounts, flow direction and demographics are outside what this sensing method can deliver, and the report states this plainly.
• Thresholds must be calibrated on-site for the actual mounting heights and geometry; lab defaults are starting points only.
• Outdoor platform deployment needs a weatherproof (IP65) enclosure and protected cable runs, which the buyer provides — the kit ships as an indoor/demo-grade assembly.
• Alerts depend on Wi-Fi; the node keeps sensing and driving the sign locally during outages, but staff push alerts pause until the link returns.
• This is an advisory aid for crowd management, not a certified safety system — it supports staff decisions, it does not replace them.
• Dense luggage trolleys and slow-moving queues can read as high density; the calibration guide documents these confounders.