Monocular Depth Estimation for autonomous navigation
Integrated and implemented the algorithm to use a single camera with reinforcement learning techniques and raspberry pi to estimate depth and ego-motion of the robot’s surrounding in order to get the robot a clearer picture of its surrounding and make the navigation well planned.
The video above demonstrates real-time monocular depth estimation from the robot.
In real time depth estimation, it determines the distance of different objects in front of the robot in form of a depth map which denotes nearer objects and its surrounding by yellow color and as the distance of objects from robot goes on increasing, the color representing depth also goes on changing gradually.
This information can be used by processing the depth maps and extracting information required for navigation.
I am still researching on how to optimize this process by implementing reinforcement learning with GPS coordinates and achieve autonomous navigation.
