Problem: Underground mine rail systems need early warning of ceiling hazards and derailment.

Approach: Built a Raspberry Pi inspection unit using dual cameras, YOLOv5n, an MPU9250 IMU, LoRa, CRC checking, and MODBUS integration.

Result: Achieved 97% ceiling-hazard accuracy and derailment classification with 96% precision and 98% recall.

Problem: UWB ranging loses accuracy in multipath and non-line-of-sight environments.

Approach: Combined Gaussian Process bias correction with a Bayesian Conv-BLSTM and spatial-temporal attention.

Result: Produced a robust LOS/NLOS mitigation method with uncertainty estimation, published in Mathematics.

Problem: Conventional shaft inspection cannot provide timely warning of deformation and alignment hazards.

Approach: Integrated CMOS cameras, LiDAR, laser collimation, and ROS-based processing.

Result: Validated deformation, perpendicularity, and vibration measurements for intelligent shaft monitoring.

Problem: Dust, humidity, and debris can make ground-level derailment sensors unreliable in mine shafts.

Approach: Deployed YOLOv5n and OpenCV on Raspberry Pi for remote visual track-state detection.

Result: Demonstrated real-time on-track and off-track classification on an edge device.

Problem: Biofloc farms require continuous water-quality monitoring and timely intervention.

Approach: Combined IoT sensors, cloud monitoring, and a decision-tree regression model.

Result: Predicted water condition with 79% experimental accuracy and delivered readings through an Android interface.

Problem: Home automation needs location-aware control without requiring manual input.

Approach: Classified room-level Wi-Fi RSSI fingerprints with a decision tree deployed on ESP32.

Result: Achieved 92% location-classification accuracy for automatic appliance control.

Problem: Waste bins need hygienic operation and timely collection.

Approach: Integrated proximity sensing, displays, environmental sensors, GSM alerts, and an online dashboard on ESP32.

Result: Delivered automatic lid control, live capacity monitoring, and full-bin notifications.

Problem: Remote sensor readings need reliable long-range delivery to industrial monitoring software.

Approach: Connected STM32 sender and receiver units through SX1278 LoRa and converted received data to MODBUS.

Result: Displayed real-time ultrasonic and potentiometer readings in Modscan.

Problem: Campus facilities needed safer, touch-free operation during COVID-19.

Approach: Combined temperature screening, contactless sanitation, smart waste handling, voice control, and smoke detection.

Result: Produced one integrated Arduino-based prototype covering five campus workflows.

Problem: Automatic gates needed a contactless way to enforce mask requirements.

Approach: Used an ESP32-CAM with a Teachable Machine classifier and Arduino-controlled indicators and servo.

Result: Allowed gate access only after a positive mask classification.

Problem: A Rubik's Cube requires coordinated physical manipulation and algorithmic move planning.

Approach: Linked a Python GUI and Kociemba solver to Arduino-controlled push, hold, and rotate mechanisms.

Result: Built a working three-servo robot capable of executing generated solutions.

Problem: Local environmental readings need to be available across connected devices.

Approach: Interfaced LDR, DHT11, and BMP180 sensors with ESP32 and a browser-based interface.

Result: Delivered live light, pressure, altitude, humidity, and temperature measurements over Wi-Fi.

Problem: Secure entry needs both identity and health-condition checks.

Approach: Combined ESP32-CAM face recognition with MLX90614 temperature sensing and Arduino control.

Result: Opened the gate only when the face was recognized and temperature was within the configured limit.

Problem: Manual bottle filling is inconsistent and labor intensive.

Approach: Built a sensor-driven conveyor with timed pump control, user-adjustable settings, and status display.

Result: Automated bottle detection, filling, transfer, and removal positioning.

Problem: A mobile robot needs simple hands-free directional control.

Approach: Connected an Android voice interface to Arduino through HC-05 Bluetooth and an L298N motor driver.

Result: Implemented reliable go, reverse, left, right, and stop commands.

Problem: Low-cost automated plotting requires precise two-axis motion from accessible components.

Approach: Repurposed DVD-drive steppers with Arduino control, a servo Z-axis, and Processing-generated paths.

Result: Built a compact CNC plotter that reproduces digital drawing instructions.

Problem: An exposed tube cutter could operate with its safety cover open.

Approach: Designed a relay, limit-switch, 555 timer, and transistor interlock with an audible warning.

Result: Prevented startup and forced shutdown when the cover was removed.

Problem: Small PCB or component bends were escaping production inspection.

Approach: Designed a BJT-based tester using fiber sensing, limit switching, lamps, and a buzzer.

Result: Detected deformations as small as 1.2 mm with clear pass/fail feedback.