Simultaneous Localization and Mapping using object detection and Reinforcement Learning for automated Robot

Considering the era of robot automation and Artificial intelligence, Autonomous robots have gained huge importance. These robots, once constructed, can be employed to do many complex tasks which reduces human labour and can be used in situations where human life is endangered. The proposed research work introduces a novel autonomous robot system, seamlessly integrating Simultaneous Localization and Mapping (SLAM) with reinforcement learning (RL) for intelligent and adaptive navigation in dynamic environments. The built system is designed with a monocular camera, employing the Oriented FAST and Rotated BRIEF (ORB) algorithm to extract key features from consecutive frames. This information is utilized to calculate the robot’s movement, providing the foundation for SLAMbased mapping. The reinforcement learning component is a key feature, training an agent to autonomously navigate the environment. The agent receives feedback in the form of positive rewards for clear paths and negative rewards for encountered obstacles. This adaptive learning mechanism ensures the robot’s ability to make informed decisions and dynamically adjust its path based on evolving environmental conditions. All these functionalities are integrated using the Robot Operating System (ROS). The proposed system effectively maps itself and builds maps of its path while avoiding obstacles coming in the path and choosing the optimal path. The system gives good results in an indoor environment. The proposed system undergoes rigorous testing in diverse scenarios, demonstrating its efficacy in autonomous navigation, obstacle avoidance, and dynamic map updating. Results indicate improved mapping accuracy and robust performance, validating the integration of SLAM and RL for enhanced autonomy. This research contributes a holistic approach to autonomous robotics, offering insights into the seamless integration of sensor data processing, SLAM, object detection, and reinforcement learning for intelligent and adaptive robotic navigation.