This week, we mainly focused on integrating our products and fine-tuning our sub-systems in preparation for the interim demo.
HojunĀ managed to finish building the gripper system and complete testing on tools after all the necessary parts arrived. He managed to pick up items ranging from 0.1kg to 1kg and 5cm to 2cm in length in over 400 tests. He also built the arm out of the two servos and plans to integrate it with the Xavier with Omar next week. A video of his tests is show below.
Omar decided on using ORB SLAM and tested it on his own robot and Jetson using a depth camera, motors, and an Arduino. He also used a rudimentary gripper to grip items like markers and pencils. For the actual robot performance, he managed to use the depth camera to integrate with the Jetson and run RTABMAP SLAM packages and manually move the wheel motors to maneuver to pick up markers at different locations. For the path planning, he decided to simply make the robot move away from walls while keep track of the closest object it has seen and move to them in ascending order of distance. If there are no more remembered objects to be pursued, it will spin and continue looking for more on its depth camera while veering away from walls. Next week he plans on integrating his path planning algo with the moving robot and Xavier with Hojun.
Andy tested different color thresholding algorithms to help improve the color detection. After testing downsampling methods, morphological filtering, and applying Gaussian Blur, he found that the blurring worked best for now and used it to further tune the best HSV range for neon pink tape. He also added task bars for simulating the bin on the robot and displayed centroids, bounding boxes, and indexing for each pink-taped object seen on the live video feed. Within the bin, these objects would be counted and the total count is displayed on the top of the screen while the contours are not drawn within the bin. (as if they have already been collected). next week, he hopes to continue fine-tune the algorithm and proceed with testing with the camera on actual pink-tape wrapped tools on a dark workshop-like floor.
As a team, next week we hope to give a good interim demo, complete the ethics assignment, and work together to begin integration of all three sub-systems to hopefully create a functional moving and autonomous robot.