For the most part, our team accomplished all of the goals we set out to do. The Baxter robot was able to play a game of checkers with little human intervention. The robot was also robust to failures -- if there are issue that arise due to failure of camera of MoveIt! the project allows for it to attempt other things until it succeeds. Overall, our project through the use of the planning, sensing, and algorithmic portions is able to consistently play an intelligent game of checkers.

After working with the Baxter robot, we felt more comfortable using a real robot to accomplish tasks. We also experienced firsthand the difficulties with working with a real system. Furthermore, as discussed in the Challenges section, we learned the frustrations of using MoveIt! that calculates inverse kinematics randomly.

Lessons

• Time management of the project
  • Booking of the robots required advance planning especially later on in the project deadline
  • One of the largest projects and required a communication and planning to get parts completed
• How to use R.O.S.
  • Got practice with R.O.S. and learned tricks for using RVIZ, launch files, etc.
  • Learned how messages and topics work
• Inverse Kinematics of checkers piece moves
  • Spent the longest time working with picking up a piece with MoveIt!

Things we would like to add later on would be to remove human input from the robot checkers game. Ideally, we would be able to run the robot and have it play games in a manner that handles exceptions and is robust. Other things we would have liked to develop further would be to remove the AR tags from the game, replacing it with computer vision. We hope to try other versions of movement packages, including re-implementing MoveIt or using CHOMP for less randomness in the path planning.