Enhancing Autonomous Navigation in Quadrupeds through RGB-D Camera Integration and Visual Waypoints
Disciplines
Robotics
Abstract (300 words maximum)
The multifaceted capabilities of RGB-D cameras present an avenue for real-time area mapping and empower robots with algorithmically-driven situational awareness. By amalgamating extracted point clouds with synchronized video and directing this composite data through detection platforms like YOLO or Faster R-CNN, we can enhance the quadrupedal Unitree Go1's prowess in persistent object recognition. This approach leverages the depth camera for autonomous pathfinding, eschewing reliance on local GPS. Instead, the robot navigates by visually establishing waypoints, drawing from both naturally occurring markers, such as landmarks, and universally recognized symbols like ArUco markers equipped with custom dictionaries. We further propose a convergence of this method with global GPS tracking, fusing expansive area mapping with real-time navigation, thereby augmenting the accuracy in both destination pinpointing and waypoint identification.
Academic department under which the project should be listed
SPCEET - Robotics and Mechatronics Engineering
Primary Investigator (PI) Name
Muhammad Hassan Tanveer
Enhancing Autonomous Navigation in Quadrupeds through RGB-D Camera Integration and Visual Waypoints
The multifaceted capabilities of RGB-D cameras present an avenue for real-time area mapping and empower robots with algorithmically-driven situational awareness. By amalgamating extracted point clouds with synchronized video and directing this composite data through detection platforms like YOLO or Faster R-CNN, we can enhance the quadrupedal Unitree Go1's prowess in persistent object recognition. This approach leverages the depth camera for autonomous pathfinding, eschewing reliance on local GPS. Instead, the robot navigates by visually establishing waypoints, drawing from both naturally occurring markers, such as landmarks, and universally recognized symbols like ArUco markers equipped with custom dictionaries. We further propose a convergence of this method with global GPS tracking, fusing expansive area mapping with real-time navigation, thereby augmenting the accuracy in both destination pinpointing and waypoint identification.