Showing comparison view.
Percentiles Key: 100th · 90th · 80th · 70th · 60th center 40th · 30th · 20th · 10th · 0th
Showing comparison view.
All data is filtered. Visualizations will stop updating until you return to the sampled space.
Displaying 0 trajectories of 0
Connect to our wildfire suppression simulator and optimizer. In this domain, rewards are given by tree harvest volumes and costs are incurred by spending money on wildfire suppression. Letting a wildfire burn may reduce the harvestable timber, but it will also reduce fuels on the landscape and fire suppression expenses.
More details on this simulator are available in the Proceedings of the IEEE Symposium on Visual Languages and Human-Centric Computing, Facilitating Testing and Debugging of Markov Decision Processes with Interactive Visualization.
This simulator is based on data produced by the researchers of a computationally expensive Markov Decision Process for wildfire suppression policy.
Since the wildfire simulator is computationally too expensive to evaluate for all the possible policies, researchers implemented a Model Free Monte Carlo with independencies (MFMCi) algorithm, which builds trajectories from a database of state transitions.
Note: Additional information on the domain, algorithm, and authors will be available after publication.
The MFMCi researchers created a toy domain to illustrate the properties of MFMC and MFMCi, then hooked the domain into MDPvis.
Note: Additional information on the domain, algorithm, and authors will be available after publication.
If you start a server running on your computer that supports CORS, you can connect this visualization to your domain without hosting the visualization code.
If you make a server publicly accessible, please let us know so we can list it here. If you need help integrating your server, please open an issue so we can help. We have documentation in the README.
This visualization allows you to explore a Markov Decision Process (MDP). The controls provide means of exploring changing reward, transition, and policy functions towards discovering bugs and understanding the system.