Automatic Generation of Object Shapes With Desired Affordances Using Voxelgrid Representation

Area of Science:

• Computer Graphics
• Artificial Intelligence
• Product Design

Background:

• Designing 3D objects is complex, with limited tools for exploring design possibilities.
• Functional requirements (affordances) are crucial for object design, dictating their action possibilities.
• The principle 'form follows function' suggests a strong link between an object's shape and its capabilities.

Purpose of the Study:

• To explore shape generation for 3D objects conditioned on desired affordances.
• To develop an algorithm for generating voxel-grid object shapes that provide specific functionalities.
• To bridge the gap between functional requirements and 3D model creation in computer-aided design.

Main Methods:

• Utilized an artificial neural network to learn a function-to-form mapping from affordance-labeled object datasets.
• Developed a method to combine forms that provide individual affordances into a single object shape.
• Implemented simulation-based affordance tests to verify the functionality of generated objects.

Main Results:

• Successfully generated 3D object shapes (voxel grids) that afford specified functionalities.
• Demonstrated the algorithm's capability to create objects with multiple desired affordances, such as contain-ability, sit-ability, and support-ability.
• Validated the generated forms through simulation, confirming they possess the intended functional properties.

Conclusions:

• The developed algorithm effectively generates 3D object shapes based on desired affordances, adhering to the 'form follows function' principle.
• This approach offers a novel method for computer-aided design, potentially accelerating the design process by automating shape generation from functional requirements.
• The study highlights the potential of AI in creating functional 3D artifacts tailored to specific needs.