My Projects

Exploring the intersection of embedded systems and artificial intelligence

Cadastral Map OCR System

Domain-Specific Document AI

Designed a domain-specific OCR pipeline for cadastral map documents using EasyOCR and OpenCV. Implemented multi-color text detection, fallback OCR strategies, and symbol recognition modules with configuration-driven detection pipelines.

Key Achievements:

  • Achieved 95% text detection accuracy and 92% symbol detection precision
  • Processed 63 annotated maps with 2-3 seconds average latency
  • Built modular, reproducible experimental framework
  • Generated structured JSON and Excel outputs for GIS analytics
EasyOCR OpenCV Python NumPy Pandas

Crowd Prediction & Safety Monitoring

LSTM-Based Temporal Prediction

Implemented YOLO-based detection with LSTM-based temporal prediction for crowd safety monitoring. Performed spatiotemporal feature extraction and behavioral pattern analysis with real-time dashboard visualization.

Key Achievements:

  • Integrated YOLO detection with LSTM temporal modeling
  • Developed Streamlit dashboard for real-time analytics
  • Implemented spatiotemporal feature extraction pipeline
  • Created behavioral pattern analysis algorithms
YOLO LSTM Streamlit TensorFlow

Robotic Localization & Control Firmware

Vision-Based Position Estimation

Integrated IMU and vision sensor data using I2C and UART interfaces for position estimation. Developed firmware routines for sensor calibration, noise filtering, and control signal stability.

Key Achievements:

  • Implemented sensor fusion algorithms for accurate localization
  • Developed calibration routines for IMU drift compensation
  • Applied noise filtering techniques for stable control
  • Integrated with ROS for simulation validation
ESP32 MPU6050 I2C Camera ROS

4DOF Robotic Arm Control System

Real-Time Servo Control

Designed firmware for real-time servo motor control with feedback monitoring. Tested actuation accuracy and resolved jitter through timing optimization and signal filtering.

Key Achievements:

  • Implemented PWM control for precise servo positioning
  • Developed feedback monitoring for position verification
  • Optimized timing to eliminate servo jitter
  • Created kinematic control algorithms
ESP32 Servo Motors PWM Embedded C

WiFi Camera Car Control System

ESP32-CAM with Advanced Web Interface

Developed an advanced WiFi-controlled car using ESP32-CAM with real-time video streaming. Features a professional web interface with glassmorphism design, WebSocket communication for low-latency control, and live camera feed with latency monitoring.

Key Achievements:

  • Built custom HTML/CSS/JS web control interface with modern UI/UX
  • Implemented dual WebSocket channels for camera streaming and car control
  • Created PWM-based speed control with real-time slider adjustments
  • Integrated LED light control with intensity adjustment
  • Achieved smooth real-time video streaming with latency display
ESP32-CAM Embedded C WebSocket Motor Control HTML/CSS/JS

Pan-Tilt Camera Control System

ESP32-CAM with Servo-Based Movement

Extended the WiFi camera car with pan-tilt functionality using servo motors for 2-axis camera movement. Enables remote directional camera control synchronized with vehicle movement.

Key Achievements:

  • Integrated dual-axis servo control for camera positioning
  • Synchronized camera movement with vehicle direction
  • Implemented smooth servo transitions for stable video
  • Extended web interface with pan-tilt control buttons
ESP32-CAM Servo Motors PWM Embedded C

Joystick-Controlled Robot

ESP-NOW Wireless Communication

Built a wireless robot control system using ESP32 and ESP8266 with ESP-NOW protocol. Implemented joystick-based control with real-time data transmission for smooth, low-latency operation.

Key Achievements:

  • Implemented ESP-NOW protocol for peer-to-peer communication
  • Developed joystick analog input processing (X/Y axes)
  • Created transmitter (ESP32) and receiver (ESP8266) firmware
  • Achieved low-latency control with 100ms update rate
ESP32 ESP8266 ESP-NOW Joystick Control

ROS2 Gazebo Multi-Robot Simulation

Hardware Validation Through Simulation

Built a ROS2 + Gazebo simulation environment that closely matched real hardware behavior to reduce trial-and-error on physical robots. Validated navigation logic before hardware deployment.

Key Achievements:

  • Created 10m × 10m environment with multiple omni-directional robots
  • Modeled sensor placement and kinematics based on actual hardware
  • Implemented payload pickup/drop-off zones and docking stations
  • Used simulation results to validate navigation logic before deployment
ROS2 Gazebo URDF Navigation Python

Self-Balancing Two-Wheeled Robot

PID Control with Real Sensor Noise Handling

Built a self-balancing robot using MPU6050 IMU and ESP32, implementing PID control tuned based on real sensor noise and drift. Demonstrates practical understanding of how control theory behaves on real hardware.

Key Achievements:

  • MPU6050 IMU calibration and drift compensation
  • Motor driver and power control optimization
  • PID tuning based on real sensor noise behavior
  • Bluetooth-based parameter tuning and interaction
ESP32 MPU6050 PID Control Bluetooth Motor Control

Offline Geo-Fencing Alarm System

ESP32-C3 + GPS + OLED | No Internet Required

Developed a fully offline boundary detection system using ESP32-C3, GPS module, OLED display, and buzzer. Emphasis on reliability without internet dependency for field applications.

Key Achievements:

  • Implemented fully offline geo-fence boundary detection
  • GPS coordinate parsing and boundary zone calculations
  • Visual feedback on OLED with audio alarm triggers
  • Low-power design for extended field operation
ESP32-C3 GPS OLED Geo-Fencing Embedded C

Health Monitoring System

MAX30102 Heart Rate & SpO2 Sensor

Developed a real-time health monitoring system using MAX30102 sensor for continuous heart rate and blood oxygen saturation (SpO2) measurements. Implemented I2C communication and signal processing algorithms for accurate readings.

Key Achievements:

  • Integrated MAX30102 sensor with I2C communication protocol
  • Implemented IR signal detection for finger placement verification
  • Developed heart rate and SpO2 calculation algorithms
  • Created serial output for real-time monitoring display
MAX30102 Arduino I2C Healthcare IoT

ESP32 Video Streaming Application

Python-Based Video Capture & Recording

Created a Python application to receive and record video streams from ESP32-CAM via serial communication. Implemented JPEG frame parsing and real-time video display with OpenCV.

Key Achievements:

  • Developed serial communication parser for JPEG frame extraction
  • Implemented real-time video display using OpenCV
  • Created video recording functionality with codec optimization
  • Built frame synchronization using JPEG markers (0xFFD8–0xFFD9)
Python OpenCV Serial Communication ESP32-CAM

Electronic Voting Machine (EVM) System

Arduino-Based Voting Solution

Built a functional electronic voting machine using Arduino with LCD display and button inputs. Features real-time vote counting, result calculation with percentages, and visual feedback using LED indicators.

Key Achievements:

  • Developed secure voting logic with button debouncing
  • Implemented real-time vote display on 16x2 LCD screen
  • Created result calculation with percentage distribution
  • Added LED visual feedback (green for vote recorded, red for results)
  • Built winner determination algorithm with tie detection
Arduino LCD I2C Embedded C Digital Logic

GPS Module Integration

Location Tracking System

Integrated GPS module with Arduino for real-time location tracking and navigation applications. Implemented NMEA sentence parsing and serial communication for geographic data acquisition.

Key Achievements:

  • Integrated GPS module with serial communication
  • Implemented NMEA data parsing for coordinates
  • Developed real-time location tracking functionality
  • Created data logging for geographic information
GPS Module Arduino UART NMEA Protocol

Research Publication

Comparative Evaluation of Communication Protocols with Vision-Based Localization

Sumukh Arood V, Dhesurya Kella, Mervin Joe Thomas

2025 International Conference on Robotics and Mechatronics (ICRM) | IEEE

Designed a decentralized wireless communication framework for real-time control of omni-directional mobile robots. Compared ESP-NOW and Wi-Fi TCP/IP protocols for latency, reliability, and power efficiency. Integrated ArUco-based vision localization achieving <0.5cm distance error and <4° yaw accuracy. Research conducted at NITK Surathkal Robotics Lab.

DOI: 10.1109/ICRM66809.2025.11349098
Published