Costas Frost

Home Featured Projects Contact

Featured Projects

Hi! I am a simulation engineer / synthetic data developer who works in Unity, Blender, Godot and various game engines. I have the privilege of working on research that pushes forward our current knowledge and use cases of AI. I presently work at MIT Lincoln Laboratory, and I get to work on a variety of interesting projects with brilliant people!



MIRAGE

Co-lead a team of developers to create a cross platform, co-located AR challenge environment. We developed the environment for Hololens 2 that tasked teams to compete in complex challenges and puzzles, where one player acting as an operator wore the HoloLens headset and communicated with a handler and their support team to solve challenges in the environment.


Check out the MIRAGE challenge



Tesse


We developed a RL training environment with a Unity-Python bridge. The avatar has multi-modal sensors, color, depth, and segmented ground truth. My role on this project was creating graphics to a high fidelity, assembling and solving technical solutions for the environments. In order to do this I created procedural tools for populating the environment and solutions for error checking. I also developed shaders for the different vision modalities. Our Reinforcement Learning training environment was featured as part of a workshop and challenge at ICRA 2020.


ICRA 2020 Workshop | Github | Paper



HAPI Lab

One of the cutting edge AI research topics of today is "Human Machine Teaming"(HMT) our vision was to create a space with a sophisticated tools and the knowledge to use them. I was part of the team that set up, integrated, and explored various sensors and scanners for real time data capture. Since then I have been involved in several HMT projects that started within the HAPI lab.


Article



Virtual Environments for Autonomous Vehicles

The goal of this effort was to create a real time reinforcement learning training environment using game engines such as Unity. This would become a valuable tool to create synthetic data sets for data starved environments. I developed techniques and tools to procedurally generate massive real time environments based on geographic locations.


Article