nova: High Performance Automated Planning

The "nova" library contains high performance implementations of MDP, POMDP, and Dec-POMDP algorithms for general-purpose use.

(This page has not yet been completed. Check back later this year.)

 
 

Reasoning Under Uncertainty

Model any multi-step decision-making problem by states (relevant configurations of the world) and actions (relevant operations that change the world). Transitions between states can be probabilistic. Reward or penalize the agent for its behavior.

Optionally, include support for observation uncertainty, and the addition of more than one agent.

High Performance Algorithms

Leverage Nvidia's GPUs to solve any problem at 10 times the speed. The library provides a variety of algorithms to employ, ranging from Perseus to LAO*.

C++ and Python

Apply the library to any C++ and Python application. Support for ROS applications is effortless and continues to be a major focus in the design. Additionally, a simple function-based interface enables many other languages to call our library.