Published Paper Details:

This paper presents the design and development of an autonomous indoor surveillance and security robot capable of reliable operation in GPS-denied environments such as offices, warehouses, residential buildings, and industrial facilities. The proposed system addresses the limitations of traditional fixed-camera surveillance by integrating autonomous navigation, real-time perception, and multi-sensor threat detection into a mobile robotic platform capable of continuous operation without constant human supervision. The core navigation capability is achieved through visual-inertial odometry, which fuses visual data from a USB webcam with inertial measurements from an MPU6050 six-axis inertial measurement unit comprising a three-axis accelerometer and a three-axis gyroscope. This sensor fusion approach mitigates the scale ambiguity limitation of monocular visual odometry and reduces cumulative drift errors associated with wheel encoder-based odometry. The fused pose output drives a continuous simultaneous localization and mapping loop, implemented using ORB-SLAM3 operating in monocular-inertial mode within the ROS 2 Jazzy Jalisco framework, that generates a persistent map of the environment and maintains self-localization throughout autonomous operation. The robotic platform employs a six-wheel drive configuration powered by six 25GA370 DC geared motors rated at twelve volts. Motor control is implemented using three TB6612FNG dual H-bridge motor driver modules. High-level computation is performed on a Raspberry Pi 4 Model B single-board computer, while an ESP32 microcontroller handles exclusively low-level real-time motor PWM control. The system integrates an HC-SR04 ultrasonic sensor mounted on an SG90 servo for dynamic obstacle scanning, an MQ-2 smoke and gas sensor for environmental hazard detection, and robotic arms implemented in single-servo and double-servo configurations using MG90S micro-servo motors controlled via a PCA9685 PWM driver. The software architecture is built on the Robot Operating System 2 Jazzy Jalisco framework, incorporating visual-inertial odometry, SLAM-based mapping, computer vision-based threat detection using OpenCV with a MobileNet-SSD v2 model deployed via TensorFlow Lite, and sensor-fusion-based obstacle avoidance. Component-level and integrated system testing was conducted in real indoor environments. The system successfully demonstrates stable multi-waypoint autonomous navigation, obstacle avoidance without halting, environmental hazard detection with hardware-level emergency motor stop, and real-time computer vision processing on embedded hardware. Navigation accuracy is expected to be within fifteen to twenty centimetres under favorable indoor conditions, consistent with reference literature on comparable low-cost sensor configurations, though this figure represents a hardware-specification-based estimate rather than a directly measured experimental result. These outcomes demonstrate the practical feasibility of an integrated, affordable, and intelligent indoor surveillance solution using commercially available embedded components.