1 of 1

demo 1 and visualizer

Motion Morph: Automated Mechanism Creation for Accessible Object Adaptation

Sanjana Satagopan, Jerry Cao, Jennifer Mankoff, Shwetak Patel

Abstract

Everyday objects have fixed motion inputs that are difficult to adapt on a case-by-case basis without engineering expertise. For example, turning a jar lid or opening a door requires grip strength that many individuals with limited hand mobility cannot perform. Through MotionMorph, we introduce an automated approach to generate 3D-printable mechanism adapters that convert an object's existing motion into a desired output motion, with applications in accessibility.

Limitations of Existing Approaches

Current tools often require engineering expertise:

    • Reprise (UIST '16): Generates 3D-printable adaptation but does not transform motion type or mechanism selection [1]
    • Roman (CHI '22): Introduces 3D-printable mechanisms for robotic manipulation, but requires expert manual selection [3]
    • MechaFormer (2024): Automates mechanism synthesis from a target trajectory curve but does not produce printable geometry [4]

From Motion to Mechanism

Mechanism library

We modeled three of the most common joint types for motion translation, adapted from Roman's mechanism taxonomy: gear and rack, spur gears, and bevel gears. Each joint is parameterized by the object's physical dimensions, ensuring the generated geometry fits the specific object rather than using generic sizing.

Opening a Jar Lid By Pulling Upward

Making the Optimal Chain

We use a simulated annealing approach by beginning with an initial random sequence of mechanisms and iteratively tweaking the chain to find the optimal system.

Start with a Random Chain

Simulate Full Motion

Score Against Target

Tweak Chain (swap, add, remove)

Visualizer

We designed an interactive 3D visualizer displays the mechanism assembly alongside motion indicators showing the input and output axes.

Digital Twins

As part of the process, we discovered how we can recreate a digital copy of the real world, building off previous research pipelines such as CARTO [5].

Future Work

Future work includes enabling users to specify a target output path as a spatial curve rather than a single axis. Additional directions include automatic motion axis detection from object scans, attachment geometry generation to integrate adapters directly onto objects, and extending the mechanism library.

Opening a Jar by Pushing

Closed

Open

References

[1] Chen et al. Reprise: A Design Tool for Specifying, Generating, and Customizing 3D Printable Adaptations on Everyday Objects. UIST 2016.

[2] Li et al. Robiot: A Design Tool for Actuating Everyday Objects with Automatically Generated 3D Printable Mechanisms. UIST 2019.

[3] Li et al. Roman: Making Everyday Objects Robotically Manipulable with 3D-Printable Add-on Mechanisms. CHI 2022.

[4] Bolanos et al. MechaFormer: Sequence Learning for Kinematic Mechanism Design Automation. arXiv 2025.

[5] Heppert et al. CARTO: Category and Joint Agnostic Reconstruction of ARTiculated Objects. CVPR 2023.

Scene

Gear and Rack Bevel Gears Spur gears

Bounding Box

Reconstruction

Open