💻 computer science

COMPAct: Computational Optimization and Automated Modular design of Planetary Actuators

This paper introduces COMPAct, a framework that computationally optimizes and automatically generates CAD models for four types of planetary actuators to minimize mass and width while maximizing efficiency, with experimental validation confirming the high performance of the resulting fabricated designs.

Original authors: Aman Singh, Deepak Kapa, Suryank Joshi, Shishir Kolathaya

Published 2026-04-03

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Aman Singh, Deepak Kapa, Suryank Joshi, Shishir Kolathaya

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are building a robot arm, like a high-tech version of a human arm. To make the arm move with the right strength and speed, you need a "muscle" (a motor) and a "joint" (a gearbox) that translates the motor's fast spinning into powerful, controlled movement.

For a long time, designing these joints has been like trying to bake a perfect cake by guessing the ingredients. You might get a working cake, but it could be too heavy, too wide, or waste a lot of energy. Engineers usually had to manually tweak the gears, draw the blueprints by hand, and hope for the best.

Enter COMPAct.

Think of COMPAct as a super-smart, automated "Robot Architect" that does three magical things:

It finds the perfect recipe: It calculates the exact size and shape of the gears to make the robot arm as light and efficient as possible.
It draws the blueprints instantly: Once it finds the perfect recipe, it automatically draws the 3D model.
It prints the cake: It sends those blueprints directly to a 3D printer, so you can build the part without ever touching a pencil or a ruler.

The Four "Gearbox Families"

The paper explores four different ways to arrange the gears, like four different families of tools, each with its own personality:

The Single-Stage (SSPG): Think of this as the sprinter. It's simple, fast, and very efficient, but it can only handle small speed changes. It's great for light, quick movements.
The Compound (CPG): This is the marathon runner. It uses two sets of gears working together. It's a bit more complex and has a little more "slop" (play) in the joints, but it can handle bigger speed changes than the sprinter.
The Double-Stage (DSPG): This is the heavy lifter. It stacks two gear systems on top of each other to get huge strength. It's powerful but a bit less efficient because there are more moving parts rubbing against each other.
The Wolfrom (WPG): This is the magician. It's a special, compact design that can squeeze massive strength into a tiny space. It's the only one that can handle the widest range of speeds, from very slow to very fast.

How COMPAct Works (The "Magic" Process)

1. The Optimization (The Brain)
Imagine you tell the computer: "I need a robot joint that is strong, light, and fits inside a specific size."
The COMPAct system then runs thousands of simulations in seconds. It tests every possible combination of gear sizes, tooth counts, and arrangements for all four gearbox families. It's like trying on 10,000 different pairs of shoes to find the one that fits your foot perfectly while being the lightest and most comfortable.

2. The Mass Model (The Scale)
To make sure the robot isn't too heavy, the system uses a detailed "digital scale." Instead of just guessing how heavy a gear is, it breaks every single part down into simple shapes (like cylinders and cones) to calculate the exact weight. This ensures the final robot won't be weighed down by its own joints.

3. The CAD Automation (The Printer)
This is the coolest part. Once the computer picks the winner, it doesn't just give you a list of numbers. It automatically builds the 3D model.

No manual drawing: You don't need to open complex design software and move lines around.
No redesign: If the computer decides the gears need to be slightly bigger, the whole 3D model updates itself instantly.
Ready to print: The result is a file you can drop straight into a 3D printer.

The Real-World Test (Did it work?)

The researchers didn't just leave it on the computer. They printed two of these robot joints:

The Sprinter (SSPG): It was incredibly efficient (60–80% efficient) and very stiff (didn't wiggle much). It was the clear winner for moderate speeds.
The Marathon Runner (CPG): It was a bit less efficient and had a tiny bit more "wiggle" (backlash), but it still worked great for its specific job.

The "3D Printing" Catch:
The paper notes that because they used 3D printers (which are great for speed and cost) instead of industrial metal machines, the gears weren't perfectly smooth. This caused a bit more friction, so the real-world efficiency was lower than the computer predicted. However, the system still found the best possible design for 3D printing, proving that even with imperfect materials, smart design wins.

Why This Matters

Before COMPAct, if you wanted a custom robot joint, you needed a team of engineers and weeks of work. With COMPAct:

Speed: You can go from an idea to a printed part in minutes.
Cost: It's cheap because it automates the design and uses 3D printing.
Performance: It ensures the robot is as light and strong as physics allows, rather than just "good enough."

In short, COMPAct turns the art of robot design into a science, letting anyone (or any robot) build the perfect "muscle" for their machine with the push of a button.

1. Problem Statement

The optimal design of robotic actuators is a critical research area, yet significant gaps remain in current methodologies:

Limited Gearbox Optimization: Existing design frameworks often optimize link lengths or spring stiffness but neglect gearbox parameters (teeth count, module, type) or treat them simplistically without considering efficiency, mass, or specific gearbox topologies.
Lack of Automation: While parametric CAD exists, no prior work has fully automated the design of planetary gearboxes for robotic actuators, requiring manual redesign for every new gear ratio or motor.
3D Printing Limitations: Although 3D-printed actuators offer rapid prototyping and low cost, most existing designs use fixed gear ratios and fail to systematically optimize for mass, efficiency, and manufacturability.
The Gap: There is a need for a unified framework that simultaneously optimizes gearbox parameters for performance (mass, efficiency, width) and automatically generates manufacturable CAD models for 3D printing.

2. Methodology

The authors propose COMPAct, a framework that integrates computational optimization with automated parametric CAD generation.

A. Gearbox Architectures

The framework analyzes four distinct planetary gearbox topologies to determine the optimal configuration for a given motor and gear ratio:

Single-Stage Planetary Gearbox (SSPG): Standard configuration (Sun input, Ring fixed, Carrier output).
Compound Planetary Gearbox (CPG): Uses rigidly connected planet gears of different diameters to achieve higher ratios in a single stage.
Double-Stage Planetary Gearbox (DSPG): Two successive planetary layers for very high reduction ratios.
Wolfrom (3K) Planetary Gearbox (WPG): A specialized topology using two concentric rings and rigidly connected planets to achieve extremely high reduction ratios in a compact volume.

B. Optimization Formulation

Objective Function: The framework minimizes a cost function ( $C_{act}$ $C_{a c t}$ ) that balances:
- Actuator Mass ( $M_{act}$ ): Calculated using a detailed geometric mass model (decomposing parts into cylinders, shells, cones).
- Efficiency ( $\eta$ ): Modeled using standard driving efficiency equations for gear meshes.
- Axial Width ( $W_p$ ): Penalized to ensure compactness.
- Gear Ratio Deviation: A penalty term for deviating from the target gear ratio.
Design Variables: A unified vector of 10 variables including sun/planet/ring teeth counts ( $N_s, N_p, N_r$ ), module ( $m$ ), and number of planets ( $n_p$ ) for up to two stages.
Constraints: The optimization enforces strict physical and manufacturing constraints:
- Geometric: Ensuring proper meshing and concentricity (e.g., $N_r = N_s + 2N_p$ ).
- Meshing: Ensuring integer tooth counts allow for even planet distribution.
- Interference: Preventing collisions between planets and the carrier.
- Diameter Limits: Constraining the outer diameter relative to the motor size.
Solver: Due to the discrete nature of gear teeth, the problem is solved via a brute-force search over feasible combinations, guaranteeing a global optimum within the defined bounds.

C. CAD Automation

Parametric Modeling: The optimization variables are linked directly to a parametric CAD environment via a text-based variable file.
Workflow: The optimizer updates the variable file $\rightarrow$ The CAD model automatically regenerates geometry $\rightarrow$ The model is ready for 3D printing.
Robustness: The system includes structural reinforcements (e.g., central bolts for sun gears, through-bolts for motor casings) to ensure 3D-printed parts are durable.

3. Key Contributions

Unified Optimization Framework: A novel solver that simultaneously optimizes four different planetary gearbox types (SSPG, CPG, DSPG, WPG) for mass, efficiency, and width across a wide range of gear ratios.
End-to-End Automation: The first framework to automate the full pipeline from gearbox parameter optimization to the generation of 3D-printable CAD models for planetary actuators.
Detailed Mass Modeling: A component-level mass estimation model that achieves high accuracy (RMS error < 0.03 kg) compared to full CAD mass properties, enabling fast optimization without running heavy CAD simulations.
Hardware Validation: Fabrication and experimental testing of two distinct actuators (SSPG and CPG) to validate the theoretical models.

4. Results

Simulation Results

Feasibility Ranges:
- SSPG: Optimal for low ratios (up to ~7.2:1).
- CPG: Optimal for mid-range ratios (~7.2:1 to 15:1).
- DSPG: Competitive in the 15:1–32:1 range.
- WPG: The only feasible option for high ratios (>32:1 up to 60:1).
Performance: The framework successfully identified that CPG offers the best trade-off for ratios between 7.2:1 and 15:1, while SSPG remains superior for lower ratios due to higher efficiency and lower mass.

Experimental Validation

Two actuators were 3D printed (using PLA on a Bambu Lab X1C) and tested:

SSPG (7.2:1, MN8014 Motor):
- Mechanical Efficiency: 60–80% (Theoretical prediction was 96%; discrepancy attributed to 3D printing resolution limits affecting tooth profiles).
- Backlash: 0.59° (0.010 rad).
- Stiffness: 242.7 Nm/rad.
CPG (14:1, MAD M6C12 Motor):
- Mechanical Efficiency: ~60%.
- Backlash: 2.6° (0.046 rad) – Higher than SSPG due to smaller effective planet diameters in the second stage.
- Stiffness: 201.6 Nm/rad.

Key Observation: While 3D-printed parts showed lower absolute efficiency and stiffness compared to CNC-machined or metal counterparts (due to material properties and surface finish), the SSPG consistently outperformed the CPG in efficiency, backlash, and stiffness, validating the framework's ability to select the superior topology.

5. Significance

Accelerated Design Cycle: COMPAct eliminates the manual, iterative process of designing gearboxes, allowing researchers to instantly generate optimized, printable designs for any motor and ratio.
Data-Driven Topology Selection: It provides empirical evidence on when to use specific gearbox types (e.g., avoiding CPG for high-ratio applications where WPG is better), moving beyond heuristic design choices.
Accessibility: By enabling the direct 3D printing of optimized actuators, the framework lowers the barrier to entry for developing high-performance robotic joints, particularly for research labs and hobbyists who lack access to CNC machining.
Foundation for Future Work: The framework sets a precedent for integrating physical constraints (mass, stiffness) directly into the generative design loop, paving the way for future iterations involving metal printing, advanced solvers (MINLP), and rotor inertia optimization.