For this project myself and my four teammates were tasked with developing a robotic system which was capable of autonomously reaming a patient's acetabulum based on positioning data determined by a pre-operative surgical plan created using a 3D-model of a patient's pelvis and a desired acetabular cup implant. It is difficult for surgeons to accurately ream a patient's acetabulum manually, leading to less than 50% of surgeries being within surgical safe zones, which could lead to many patients dealing with hip dislocations and discomfort. As such, the robotic system we developed was focused on improving the reaming accuracy.

I took the role of a mechatronics engineer for the team, focusing on the development of an end-effector that we could fit onto the end of a Kinova Gen-3 robotic arm and control to accurately ream the acetabulum. I designed a novel linearly-actuated design which worked by utilizing a ball screw to linearly move a carriage adapting the reaming motor and reaming shaft forward. This end-effector was also held at an angle to the robot arm and included several bearings and couplings which helped to minimize vibrations during a procedure. I led the design, prototyping, and manufacturing of this end-effector, partnering with Xometry in order to develop aluminum parts for the structural components for the final design. I then collaborated with the team to develop a cover for the overall end-effector, leading to a final product which looks as professional as it is functional.

I also took the lead in building the electrical system to power the end-effector allowing it to interface with our main ROS architecture. Utilizing an Arduino Mega, two Cytron MD30C motor controllers, an ACS712 current sensor for indirect force measurements, and limit switches, myself and a teammate were able to develop a system and Arduino script which could receive commands from ROS and apply PID admittance and velocity controls to ream to a specified depth with less than 0.1 mm error and without exceeding a force threshold of 100 Newtons.

I also took a role in verifying the accuracy to which we were reaming our pelvises to. To do this myself and a teammate used a Creaform HandyScan to create an STL file of a sawbone pelvis before and after a reaming procedure. Utilizing knowledge of the desired position of the acetabular cup for each trial, we utilized a combination of Meshmixer, Blender, and CloudCompare to determine the final accuracy of our system, which was found to have an average error of less than 1 mm across all trials.

Throughout my time working on this project I did not work on solely the mechatronic design of the system, as I helped to plan the system architecture for the project, developed project management skills from using Confluence and Jira, learned about how to set-up and utilize a ROS workspace, learned about fiducial tracking and modern surgical perception systems, helped to design the user interface, and learned about the modern hip and knee replacement procedure.

A link to the teams website can be seen below as well as a copy of our final report.

https://mrsdprojects.ri.cmu.edu/2022teamc/

For this project myself and four teammates were tasked with implementing a decomposition-free coverage planner in ROS which worked with an Ackermann steering robot. As part of the team I focused on setting up the ROS environment primarily working with teammates to set up the local planner and mapping sub-systems as well as an Ackermann steering package. I focused on optimizing the performance of these sub-system extensively, verifying that we could successfully pass movement commands along a ROS topic to the Ackermann robot which it could follow. I also aided my teammates extensively, helping with developing the architecture for our code and debugging the motion planners. Through this project I gained a lot of experience working with ROS, Gazebo, and Rviz and deepened my understanding of A* motion planning.