Perception System for Autonomous Robots Using Depth Camera

Abstract

As autonomous robots to move safely and effectively in real-world settings, they need to have an accurate sense of their surroundings. The creation of a perception system utilising an Intel RealSense RGB-D depth camera combined with inertial sensors is presented in this work in order to facilitate autonomous navigation and dependable environmental comprehension. In order to accomplish real-time object detection, distance measurement, and size estimation of nearby obstacles and targets, the suggested system records synchronised colour and depth data. When using depth information instead of traditional monocular vision systems, spatial accuracy is improved. An Inertial Measurement Unit (IMU) with a gyroscope and accelerometer is used to evaluate robot orientation, angular velocity, and movement dynamics in order to improve motion awareness and stability. Robust perception during dynamic motion and changing illumination conditions is supported by sensor fusion between visual input and inertial measures. The system creates useful spatial information needed for navigation and obstacle avoidance by processing sensor data using computer vision and geometry algorithms. The robot is able to make well-informed navigation decisions because to experimental results that show precise object localisation, dependable depth estimate, and seamless orientation tracking. For autonomous robotic applications including indoor navigation, surveillance, and service robotics, the suggested architecture offers an affordable and scalable solution.