AMD GPU-accelerated design tools for creating novel lifeforms

Project Information

training
Project Status: Complete
Project Region: Northeast
Submitted By: David Matthews
Project Email: jbongard@uvm.edu
Project Institution: University of Vermont
Anchor Institution: NE-University of Vermont
Project Address: Vermont

Mentors: Keri Toksu
Students: David Matthews

Project Description

While automatically designing machines in silico and manufacturing them as robots using 3D printers is now well established, automatically designing and instantiating living systems was only recently demonstrated to be possible (Kriegman et al., PNAS 117: 4, 2020). Novel lifeforms were assembled according to a computer-generated blueprint which tells a microsurgeon or 3D bioprinter, at some resolution of detail, precisely where all the tissues should go in relation to each other to achieve some desired behavior.

The software used to design this artificial life simulates living tissues and cells as collections of elastic voxels, the dynamics of which can be calculated concurrently on GPUs. This allows for orders of magnitude greater resolution of biological detail than the original CPU-based simulation used by Kriegman et al. However, the GPU based simulation is currently limited to Nvidia’s CUDA platform and therefore confined to Nvidia GPUs. Thus, this project aims to expand the number and kind of computers in which artificial life can be designed.

Additional Resources

Launch Presentation:
Wrap Presentation: 2

Project Information

training
Project Status: Complete
Project Region: Northeast
Submitted By: David Matthews
Project Email: jbongard@uvm.edu
Project Institution: University of Vermont
Anchor Institution: NE-University of Vermont
Project Address: Vermont

Mentors: Keri Toksu
Students: David Matthews

Project Description

While automatically designing machines in silico and manufacturing them as robots using 3D printers is now well established, automatically designing and instantiating living systems was only recently demonstrated to be possible (Kriegman et al., PNAS 117: 4, 2020). Novel lifeforms were assembled according to a computer-generated blueprint which tells a microsurgeon or 3D bioprinter, at some resolution of detail, precisely where all the tissues should go in relation to each other to achieve some desired behavior.

The software used to design this artificial life simulates living tissues and cells as collections of elastic voxels, the dynamics of which can be calculated concurrently on GPUs. This allows for orders of magnitude greater resolution of biological detail than the original CPU-based simulation used by Kriegman et al. However, the GPU based simulation is currently limited to Nvidia’s CUDA platform and therefore confined to Nvidia GPUs. Thus, this project aims to expand the number and kind of computers in which artificial life can be designed.

Additional Resources

Launch Presentation:
Wrap Presentation: 2