This was my final project for the course Mechanics of Manipulation (16-741) at CMU.

 

The project aims to explore a number of real-world manipulation tasks using analytically determined grasps and coordinated motion to move objects in a desired way. The tasks include picking and perturbing a pill cup, picking a pill cup and rotating it in-hand, and unscrewing the cap of a water bottle. The platform used is a 2x2 compliant delta robot array with each delta made out of a 3D printed TPU and driven by 3 prismatic actuators. The soft design yields advantages in both safety and dexterity which are beneficial in a human-robot interaction (HRI) context. The manipulation problems are selected from common tasks used in a medical/assistive context to show that an end effector configured in this way has viability in this space.

​

During the project I worked to develop a better synchronized control scheme for the servos in order to avoid crushing/dropping objects. This was done using Arduino code deployed onto a Wifi-enabled microcontroller. The microcontroller would recieve workspace trajectories for each 'finger' from a python server using protobuf. Each task trajectory was constructed by first picking desired object waypoints, then determining a force closure grasp using convex hull, and finally converting world space coordinates to configuration space coordinates that could be sent wirelessly to the robot.