Michael Folmer | Mechanical & Robotics Engineering

Guardian Brace V2

Powered orthotic, built full stack, independently since May 2026

System spec
Actuator	CubeMars AK60-6 KV80, MIT mode
Bus	CAN / TWAI, torque only frames
Controller	Arduino Nano ESP32
Sensing	ICM-20948 9-DOF IMU + encoder decode
Torque	±12 N·m capability, capped in firmware
Assist profile	3° deadband, ramp to 20°, hold to 30°, taper to 40°
Interface	WiFi SoftAP + WebSocket, BLE UART, SSD1306 OLED
Electronics	Custom PCB, in-house Gerber release
Structure	Creo assemblies, printed & machined mounts

Strapping in at the bench

A powered knee brace for ACL rehabilitation, assisting recovery through controlled, actuated joint motion. A brushless actuator drives the joint over CAN while a 9-DOF IMU and decoded encoder telemetry track joint angle and velocity.

I chose a CubeMars AK series quasi-direct-drive motor for its low impedance. Rehab loading needs the joint to stay backdrivable so the device can assist a patient without fighting them. High reduction geartrains give torque but resist the wearer, which is the wrong trade for therapy.

Safety drove the whole design. The firmware runs a strict passive measure, confirm, then assist sequence. On power up it transmits nothing and passively decodes encoder broadcasts to learn the wearer's range of motion. It only starts driving torque once the operator explicitly confirms.

Why the gain bytes are zeroed

Assist goes out as torque only CAN frames with the gain bytes zeroed. Two MIT byte layouts exist in the field, and a frame with bytes 0 through 5 zeroed decodes as zero stiffness and zero damping under both. An ambiguous motor firmware can't turn a damping request into a velocity slam.

Torque is hard capped, overspeed trips need consecutive confirming samples so one bad reading can't trip them, and the watchdog is sized to the motor's real broadcast period.

The wearer runs it from a phone over the brace's own WiFi access point. The microcontroller serves a self contained web panel that streams live telemetry over WebSocket, with an on board OLED and a BLE link as fallbacks.

Guardian Brace on the bench with the actuator, control board and the phone control panel — Bench setup with the phone panel

Guardian Brace custom PCBs fresh from fab — Custom PCB

Assembled Guardian Brace control board on its printed mount — Assembled board

Creo Parametric
PCB design
C++ / PlatformIO
CAN bus
Embedded control
IMU sensing
Safety architecture

More major builds

Systems carried from requirement through CAD, fabrication, electronics, firmware and test.

Two-stage variable torque actuator

Senior design capstone, proprietary

An actuator whose output torque and speed can be varied through a two-stage transmission, instead of trading motor efficiency against one fixed reduction. It grew straight out of the brace work: a rehab joint needs to move quickly and freely during gait, then hold and push hard during therapy, and one fixed ratio can't serve both well.

The work covers transmission architecture, the ratio switching mechanism, and design for manufacture of the printed and machined parts, modeled in Creo as a full parametric assembly so stage ratios and envelope can be re-solved as the design converges. The switching mechanism itself is redacted pending IP and held back from public description.

Mechanism details withheldCommercialization pending. Full design, analysis and hardware are available to discuss directly or under NDA.

Design parameters
Type	Two-stage variable torque transmission
Key mechanism	withheld from publication
Engagement	withheld
CAD	Creo Parametric, full parametric assembly
Scope	Architecture, mechanism design, DFM, prototype
Program	Senior Design II, Stevens
Status	Proprietary, commercialization pending

20:1 two-stage planetary gear drive

Transmission for a NEMA 17 stepper

A compact two-stage planetary reduction delivering 20:1 from a standard NEMA 17 stepper, designed, analyzed and bench validated end to end. Both stages share a 54 mm ring pitch diameter so the gearbox packages into a single bore, with different modules per stage (0.75 input, 1.2 output) to put the bigger teeth on the higher torque stage.

I ran Lewis bending hand calcs on the critical stage-two sun gear and checked them against static FEA in Creo. They disagreed by 69%, which I documented rather than papered over. Iterating face width raised the analytical safety factor from 0.86 to 1.72. I also built and debugged the test bench myself, tracing an intermittent fault to a floating high impedance driver node.

Drive spec
Reduction	20:1, two planetary stages
Input	NEMA 17 stepper
Ring pitch dia.	54 mm, common to both stages
Modules	0.75 input / 1.2 output
Analysis	Lewis bending vs. static FEA (Creo)
Safety factor	0.86 → 1.72 after face width iteration
FEA variance	69%, documented for follow-up
Test bench	A4988 + ESP32, 1.5 A/phase

Gearbox test bench with stepper driver wiring and printed carrier

Printed planetary stage with green planet gears in a blue ring gear

Assembled gearbox housing mounted on the NEMA 17 stepper

DIY drone build

Full mechanical and electrical integration

A drone built from nothing: airframe, wiring harness and flight firmware, taken from loose components to synchronized motor output. I modeled the assembly, printed and iterated the airframe for fit and stiffness, soldered the power and signal harness, and wrote and debugged the control code.

It never flew properly. The motors ran synchronized, but structural stiffness and motor alignment stayed the limiting factors. It's here because the wiring, the bring-up and the failure analysis are what I actually learned from.

Build spec
Airframe	3D printed, iterated for fit and tolerance
Propulsion	Brushless motors + ESCs
Power	Custom soldered power and signal harness
Control	Embedded firmware to synchronized output
Outcome	No stable flight; stiffness and alignment found as limits

Printed quadcopter frame with props and electronics on a cutting mat

Drone wiring and flight electronics during bring-up

Smaller builds

Shorter cycles, real hardware. Several taught me more by failing than by working.

Arduino table sweeper

A sweeping robot run by a multi state finite state machine handling sweep sequencing and edge detection, with an LCD reporting live status and mode.

Watch the demo on YouTube

FSM design, embedded C, mechatronics

Autonomous navigation robot

Team robot that follows a mapped path while avoiding obstacles. I designed the chassis and drive layout, the sensor and electronics architecture, and the exploded view drawing.

Robotics, CAD, electromechanical integration

Exo boot load frame

A frame to carry armor load into the ground through a boot. Tested under real loading, it failed: too much deformation, not enough stiffness, a load path that didn't do its job. Finding those limits was the result.

Load path design, failure modes, prototyping

Newton extension machine

A cable and pulley resistance mechanism that isolates the triceps while keeping joint load low, with adjustable geometry, pulley force analysis and FEA load cases.

Mechanism design, pulley analysis, ergonomics

Research & analysis

17-4 PH lattice fatigue research

Vibrational fatigue in FFF printed 17-4 PH stainless lattices for defense and aerospace use. I designed gyroid, cubic and octet specimens, ran hammer modal and piezo shaker tests, and analyzed the signatures for fatigue onset and failure modes. Led under Professor Choi at Stevens.

Modal analysis, materials characterization, research lead

RSM 3D print strength optimization

A response surface methodology experiment to raise printed part strength: isolated the highest impact print parameters, cut the non essential variables, and found the best settings from the data.

Experimental design, RSM, materials behavior

Hoeken linkage motion study

Simulated a Hoeken straight line linkage to verify the straight portion of the coupler path, plus point and distributed load studies on the links.

Kinematics, motion simulation, load analysis

What I work with

Design & CAD

Creo Parametric
SolidWorks
Parametric assembly modeling
Design for manufacture
Mechanism & gear design

Analysis

Finite element analysis
Nastran (Femap / NX)
SolidWorks Simulation
Lewis bending stress
Modal & vibration analysis
Design of experiments (RSM)
Kinematics & dynamics

Electronics & firmware

Embedded C / C++
ESP32, PlatformIO, Arduino
CAN bus / TWAI
I2C sensing, IMU integration
PCB design & Gerber release

Fabrication

FDM 3D printing
Print process & orientation tuning
Soldering & wiring harnesses
Bench test rig construction
Hardware bring-up & debug

B.E. Mechanical EngineeringStevens Institute of Technology, expected May 2027

M.Eng. RoboticsStevens Institute of Technology, expected May 2028

MichaelFolmer