AR4 Robotic Arm

Unlike the other projects I worked on at TVS, this was not my idea. Two other colleagues who had a background in robotics suggest the idea of a pick and place robot, that would sort items coming in on a conveyor belt. The robot we chose was very affordable as a proof of concept (PoC) with plans to scale up upon its success.

Once the arm was built it was handed off to me to develop the software that would compete the pick and place. The arms schematics were created by Annin Robotics, who kindly includes some useful, albeit limited software for pick and place already. While simple it would be more than sufficient for a PoC, which you can see in the gallery.

The Arm has more than proved good enough for a proof of concept! However, I discovered the speed of the arm, and its visual processing were limited by the default software, so I looked into potential improvements. I discovered somebody had made a Driver that allowed the arm to be used with the Robot Operating System (ROS2). I wanted to use this for the robot's locomotion, and a custom CNN for object detection as the default one used simple greyscale pattern matching, so was very easily confused if shapes were similar.

I made some simple scripts that allowed me to control the arm with ROS2 and was making headway in developing a model to recognise a variety of objects in the warehouse. Unfortunately, this work would never make it to a finished PoC stage. TVS SCS did not meet the profit quota for the quarter, this meant that some budget cuts had to take place. As the team I worked on was R&D focussed, the team was to be cut from 8 people to 3. Having started sooner than the others, with less knowledge of the business, I was made redundant.