zhennan-wu / rddlsim Goto Github PK

Implements a parser, simulator, and client/server evaluation architecture for the relational dynamic influence diagram language (RDDL) -- pronounced "riddle". RDDL is intended to compactly support the representation of a wide range of relational MDPs and POMDPs and support the efficient simulation of these domains. RDDL is used in a number of past and present International Probabilistic Planning Competitions (IPPCs):

• IPPC 2011 (Discrete)
• IPPC 2014 (Discrete)
• IPPC 2015 (Continuous) -- cancelled
• IPPC 2018 (Discrete and Continuous)

• RDDL Language Guide

  Please cite as

   @unpublished{Sanner:RDDL,
      author = "Scott Sanner",
      title = "Relational Dynamic Influence Diagram Language (RDDL): Language Description",
      note = "http://users.cecs.anu.edu.au/~ssanner/IPPC_2011/RDDL.pdf",
      year = 2010}

• RDDL Tutorial Slides

• RDDL Tutorial Website -- a step-by-step guide to building the Wildfire domain, simulating it in RDDLSim, and using the PROST planner with it.

• Tensorplan for RDDL -- a Tensorflow-based planner for continuous RDDL domains

When you checkout the code, the first file you'll want to look at is INSTALL.txt... this includes everything you need to start simulating, visualizing, and translating domains in < 5 minutes!

If you want to use an MDP or POMDP planner in conjunction with RDDL, please check out planner releases at the above IPPC competition web pages... all support the RDDL Client/Server protocol.