Welcome to the worlds highest fidelity simulation environment. The Simulacrum was developed from the foundation up to train autonomous agents that will eventually interact in real world space. With a primary forcus on sim-to-real transfer learning, the Simulacrum provides deep reinforcement learning agents a playground to encounter situations that may be dangerous or costly to encounter in the real world.

High Fidelity

Randomized and Procedural

Automated End-to-End

The Simulacrum is a complete technology stack dedicated to providing agent based training environment simulation. It isn't just another physics sandbox for pre-training, its a whole package pipeline for automated autonomous policy generation from code to training to inference on a physical platform. Using the latest simulation technology and modern deep reinforcement learning algorithms the Simulacrum provides a powerful foundation for any autonomous project.

Fast, Cost-effective, Safe

Explore edge cases with no risk

Depending on the sample efficiency of the method used, modern machine learning algorithms often require up to billions of samples in order to converge to an optimal solution. As such, the ability to collect that data quickly is paramount. One of the most beneficial properties of a simulation is the ability for it to be run at a speed often orders of magnitude greater than that of the physical world. In addition to this increase in speed, simulations can often be run in parallel, allowing for orders of magnitude greater data collection than real-time serial experience in the physical world. The faster such algorithms can be trained, the greater the speed of iteration and experimentation that can take place, leading to faster development of novel methods.

Welcome to the Simulacrum

Back to the top