Mr. DartBoi's head fell off after an intense shooting session.
There were many challenges we had to overcome throughout this project.
Mechanical:
Friction and fits between PLA-printed parts
Reliable trigger mechanism actuated by weak gripper actuator
Software:
Optimal path planning required tuning the number of attempts the MoveIt commander should perform for an effective but safe path.
Discrepancies in transformations over different attempts and even over the same attempt with inconsistent circle readings at a longer range.
Characterization of the dart's launch trajectory to find the reference position of the launcher for a safe and accurate launch.
We have accomplished many of our initial design criteria:
Completed a custom dart-launcher working with Sawyer robot.
Achieved stationary dartboard tracking with OpenCV's Circle Hough Transform.
Effective path-planning with MoveIt Commander.
Effective and safe control of the Sawyer robot using PID controller.
Accomplished over 90% accuracy (hitting the dartboard counts as a 'hit').
For a more effective implementation of DartBoi v2:
Make the dart-launcher with stronger material to account for the deformation of launcher over time from loading/unloading.
Better characterization of mechanical offsets (camera tilt, launcher velocity, gripper twist).
Optimal path planning is difficult as the preset 5000 attempts are not always enough. Although we use parallel processing of MoveIt commander to speed up the 5000 attempts, it still takes up to 5 seconds. Moreover, if the shortest path planned by the MoveIt commander is deemed not plausible or safe by the user, the path planner has to undergo another 5000 attempts until it finds a desirable path. Moreover, speeding up this path planning phase is essential for our next objective of hitting a moving target.
A reliable and fast method for tracking moving dartboard with predictive modeling of the target.
Faster trigger method to hit moving targets.