Result

Group 1 - Portrait of Tom Hardy

Group 2 - Portrait of a beautiful lady

Group 3 - Portrait of Prof. Shankar Sastry

Conclusion

In general, we finished our main goal and we have got portraits with relatively high quality. The main characteristics are well-expressed and recognizable. However, we are having some unstable parts in our project. Locating the height of the canvas is one of them(as this needs errors < 1 mm). If we can use the force detecting sensor, once the pen is pressed on the table, the sensor would tell the controller to stop going down anymore. Combing the AR tag and force sensor would be a better solution to this problem. Another problem is that although we made a pen holder and can precisely find the location of it, we still had a hard time to grasp it tightly using the "close" command for sawyer's gripper. This is because the holder cannot exert force on all four directions(at most 2). And we could only use the tape to tie the marker pen on the gripper. To finish this target, a more complex machine structure is needed, such as add springs, nuts and bolts.

The future work that we mostly want to finish is multi-arm drawing. First we can start from baxter, and use its both arms to draw simultaneously. And later on, we can use multiple robots to draw together. This is exciting but hard both in implementation and hardware (as we cannot move the position of the robots and it's hard to order multiple sawyer/baxter at the same time). To finish this, we need to divide the planning for each arm, and we have to avoid the collision(as the arm is pretty big), i.e. two arms have a intersect in configuration space(include physical volume) at the same timestamp(if so, one have to wait ahead). For multiple robots, coordinate calibration would also be a hard part, as the precision is required. But this could be super-cool.