Monolithic Systems Lab: Undergraduate Researcher
Soft robotics research at UIUC on continuum compliant manipulators. I redesigned the manipulator's disk assembly to cut reassembly from 10 to 3 minutes per link, and I build the 3D-printed test objects and obstacles the lab uses to benchmark grasps.
- Role
- Undergraduate Researcher, Soft Robotics
- Timeline
- Sep 2026 – Present
- Status
- Ongoing
- Stack
- SolidWorks, 3D Printing (PLA)
Overview
The Monolithic Systems Lab at the University of Illinois, led by Prof. Girish Krishnan, studies soft robotic systems. The part I work on is continuum compliant manipulators: arms that bend smoothly along their length instead of rotating at a handful of rigid joints. The question we're asking is how the way you actuate different parts of the system changes what the arm can reach and manipulate, and how to make that control as simple as possible.
The manipulator is a stack of disks threaded onto a flexible backbone, about 9 to 10 links long right now. A single tendon (a fishing line) runs the length of the backbone to a motor at its base, and pulling it bends the whole arm. Each disk also has its own servo that rotates it on the backbone. Because the tendon is shared, the orientation of each disk decides how that section bends when the line is pulled. Setting the disk orientations is effectively how you shape the arm's path toward a target.
My Role
I'm an undergraduate researcher working under PhD student Sabyasachi Dash and Prof. Krishnan. My work is on the mechanical side of the manipulator and its test setup. I design the hardware in SolidWorks, print and assemble it, and build it so the lab can run experiments faster.
So far that has meant two things. I redesigned the disk assembly so links are quick to build and replace. I also designed and printed the benchmark objects, connectors, and obstacles used in grasp and failure-point testing. My current focus is a proof-of-concept actuator that could replace the servo on each disk.
Technical Approach
Redesigned the disk assembly of the continuum compliant manipulators in SolidWorks, cutting reassembly from 10 to 3 minutes per link (70%) and letting the lab swap in new units between grasp tests.
The original disk used a lot of screws, and many of them were hard to line up and hard to reach once the rest of the arm was assembled around them. I reworked the disk around assembly access. I moved the fastener locations to where they could be reached with the stack built up, relocated the servo, and simplified the disk geometry so fewer features had to line up at once. The result goes together in a fraction of the time, which matters in two ways. A link damaged during testing can be swapped out without slowly taking apart its neighbors, and building a whole new manipulator for another test went from a long job to a quick one.
Designed and 3D printed size-graded benchmark objects, connectors, and obstacles for grasp and failure-point testing.
To measure what the manipulator can actually do, the lab needs repeatable things for it to reach, wrap around, and work past. I designed a set of size-graded objects to find where grasps start to fail, connectors for the test setup, and three different obstacles that test whether the arm can manipulate around them properly. Everything is printed in PLA, so a broken or modified part can be reprinted and back in the setup quickly.
Challenges
The main constraint on the disk redesign was access. In a stacked assembly, each disk sits between its neighbors on the backbone, so a fastener that's easy to reach on a disk by itself can be nearly impossible to reach once the arm is built. Making reassembly faster depended less on the part and more on the order you build the stack in: where your hands and a driver can still get to at each step.
The other problem is the servo itself. Each disk carries a servo, and that's a large mass mounted on every link of what's supposed to be a light, compliant structure. It throws off the dynamics of each disk and makes the arm hard to model. It's difficult to calculate and predict in simulation how the real manipulator will behave, which works against the lab's goal of making the system simple to control.
Current Status & Next Steps
The redesigned disks and benchmark objects are in use for the lab's testing. I'm now working on a proof-of-concept design that drives the backbone disks with a thermal coil instead of a servo. Running current through the coil would actuate the disk and control its orientation, without the weight of a motor on every link. If it works, the disks get lighter and their dynamics get simpler, so the manipulator should be easier to simulate, predict, and control.