SHANEL’S STATUS REPORT 3/7
For the OCR algorithm, I got Tesseract set up and working on input image files. I also wrote a python script to start taking input from the Raspberry Pi Camera and process optical character recognition, but haven’t been able to test it. The process to download Tesseract and OpenCV onto the Raspberry Pi took longer than expected, but we finally got it.
I also came up with a general navigation planning flow using a Global and Local SLAM subsystem modeled after Google’s Cartographer SLAM ROS package. Local SLAM and Global SLAM. Local SLAM will be using sensor data from the spinning lidar and camera to construct a series of submaps. Each individual map will have a limited view of each sensor at one instance in time, and will eventually be joined together by Global SLAM to form the resulting overall map. The Global SLAM subsystem is able to rearrange submaps between each other to create structure from motion, eventually yielding a complete view of the traveled area.
As a MVP, I think we will start out with a simple navigation flow:
System starts when robot is at dock
While loop: (there is unexplored frontier and/or robot is not back at dock)
- Take 2D scan from Lidar
- Find all possible unexplored frontiers
- Choose direction where depth is furthest. In the case of a tie, choose arbitrarily.
- Move robot
- Repeat loop
Exits while loop when everything has been explored and the robot is not back at dock.
I’m slightly worried about the navigation algorithm since I only have a conceptual grasp of it. Theoretically, it should be able to map the entire room within the allotted requirements. However, we have not tested the granularity and accuracy of the A1 Lidar. Next steps are to download the Cartographer ROS package, which we had some trouble doing due to unresolved dependencies and try out an initial algorithm. I also plan to start on the door detection algorithm this week. I plan to have a working first model by the end of next week. Since I won’t be in Pittsburgh, I won’t be able to test anything on hardware, but Aditi will be able to.