COMPUTER VISION AND SIMULATION
Simulation for Product Design Iteration
An agricultural technology client, had some ideas about how game technology could help them automate research and design of their automated robotic equipment but didn’t have the expertise in
house to develop and prove them out.
They were hoping to be able to replace expensive and risky testing in the field with testing in simulations to be able to narrow down the best way to set up their platform for different combinations of
crops at different stages of growth, plants of different kinds and densities. They were specifically interested in being able to generate synthetic imagery simulating what their machine vision systems would see when mounted on their platform at different heights and angles.
FS Studio not only delivered a system that produced the desired synthetic imagery, but along the way the simulation drew interest from other groups within the company for use as a tool to visualize how the system worked for potential customers snd for training on how the system worked for employees within the company.
One of the bigger challenges in creating the simulation was achieving the high crop and weed densities encountered in the real world while still delivering acceptable performance and simulation fidelity. By using the latest DOTS (Data Oriented Technology Stack) technology from Unity technologies we were able to achieve acceptable simulation performance with high resolution simulated imagery.
To produce the simulation required helping the client source appropriate 3D models of many types of plants and weeds, preparing the models to ensure fidelity and uniformity, creating texture atlases for performance, and sourcing, reworking and animating a 3D model of their robotic platform. Custom shaders were written to achieve acceptable performance when creating the synthetic imagery.