Scorpio is a passively-balanced robotic manipulator developed by Apptronik. Springs embedded in the robot compensate for the weight of successive linkages and a given payload in all configurations. As part of my graduate research, I developed a controller on this system for collaborative assembly tasks.
Prior to my involvement, the robot was capable of scripting to joint states and Cartesian poses:
The Robot Operating System (ROS) was used to control Scorpio. ROS is an open-source project connecting the global robotics community to control a wide variety of robots and interface with all kinds of sensors and peripheral hardware.
I made a ROS package that "jogs" the robot, taking Cartesian velocity inputs and streaming joint state commands to the robot using an inverse Jacobian method. Scorpio has 7 degrees of freedom, one more than is required for a given Cartesian pose. This redundancy was used to stay away from joint limits while jogging. Here I am using a 3D mouse to teleoperate Scorpio:
My research focused on improving precision insertion tasks with heavy payloads. A bowling ball was used as the payload, with a raw spaghetti noodle at the base as the insertion member. These objects were chosen to model a given application while conveying the precise forces involved with the task.
My controller compensated for the weight of the payload and actively removed a portion of its inertia. Operator interaction forces were detected with a Force/Torque (FT) sensor, and an admittance control law was used to convert these measurements to jogger inputs.
To validate this system in the collaborative task, an experiment was conducted to compare Scorpio against passive systems or without any augmentation. I ran a pilot study with 6 untrained operators, challenging them to blindly insert the fragile pin into a hole at the base of a hemispherical cup.
In the recording below I am performing the experimental task, augmented by Scorpio using my collaborative controller. On the right, the interaction forces are shown in the standard ROS visualization package:
This pilot study was small in terms of data yet useful to gather operator feedback for future controller development. The preliminary results are promising, showing an increase in operator success and a decrease in time required to complete the task.
My work with Scorpio is more thoroughly covered in my Master's thesis - Manipulator Control in Collaborative Assembly. For now, I am off to work at Los Alamos National Lab, the funding agent for my graduate studies.