EvoPoli is a search-based software engineering approach and tool that employs multiobjective optimisation genetic algorithms to automate the synthesis of approximate Pareto-optimal policies for MDPS with complex combinations of constraints and optimisation objectives.
EvoPoli is the outcome of the research paper "Evolutionary-Guided Synthesis of Verified Pareto-Optimal MDP Policies" published at the 36th IEEE/ACM International Conference on Automated Software Engineering
EvoPoli is a Java-based tool build on top of EvoChecker and uses Maven for managing the project and its dependencies, and for generating the executable jars.
-
Import the project in your IDE of preference
-
Set the following environment variable (In Eclipse go to Run / Run Configurations / Environment tab / New)
OSX: DYLD_LIBRARY_PATH = libs/runtime
*NIX: LD_LIBRARY_PATH = libs/runtime
-
Specify the configuration parameters in file config.properties
-
Run MainMDPSynthesis