Solver Documentation

Documentation for the SARSOP.jl user interface.

Contents

Index

Solver

SARSOP.SARSOPSolverType
SARSOPSolver

Base solver type for SARSOP. Contains an options dictionary with the following entries:

  • 'fast': use fast (but very picky) alternate parser for .pomdp files
  • 'randomization': turn on randomization for the sampling algorithm
  • 'precision': run ends when target precision is reached
  • 'timeout': [sec] If running time exceeds the specified value, pomdpsol writes out a policy and terminates
  • 'memory': [MB] If memory usage exceeds the specified value, pomdpsol writes out a policy and terminates
  • 'trial-improvement-factor': terminates when the gap between bounds reaches this value
  • 'policy-interval': the time interval between two consecutive write-out of policy files
  • verbose: [true/false] whether to print output from the solver. Default: true
source

File manipulation

SARSOP.POMDPFileType
POMDPFile

Writes a pomdpx file from a pomdp defined using the explicit interface of POMDPs.jl

Constructors

  • POMDPFile(filename) just stores the file name if the file exists
  • POMDPFile(pomdp::POMDP, filename="model.pomdpx"; verbose=true) writes a .pomdpx file from the pomdp definition
source
SARSOP.load_policyFunction
load_policy(pomdp::POMDP, file_name::AbstractString)

Load a policy from an xml file output by SARSOP.

source

Simulation and Evaluation

SARSOP.SARSOPSimulatorType
SARSOPSimulator

simulate a sarsop policy from the xml files

  • sim_len number of steps to use in simulation (default 10)
  • sim_num number of simulations
  • fast use fast (but very picky) alternate parser for .pomdp files
  • srand set the random seed for the simulation
  • output_file where to write the results
  • policy_filename policy file location
  • pomdp_filename pomdpx file location
source
POMDPs.simulateFunction
POMDPs.simulate(sim::SARSOPSimulator)

Simulate a SARSOP policy according to the config specified by sim::SARSOPSimulator

source
SARSOP.SARSOPEvaluatorType
SARSOPEvaluator

simulate a sarsop policy from the xml files

  • sim_len number of steps to use in simulation (default 10)
  • sim_num number of simulations
  • fast use fast (but very picky) alternate parser for .pomdp files
  • srand set the random seed for the simulation
  • memory [MB] No memory limit by default. If memory usage exceeds the specified value, the evaluator will switch back

to a more conservative (and slow) methods.

  • output_file where to write the results
  • policy_filename policy file location
  • pomdp_filename pomdpx file location
source
SARSOP.evaluateFunction
evaluate(ev::SARSOPEvaluator)

simulate a SARSOP policy according to the configuration specified by ev::SARSOPEvaluator

source

Policy Graph Generation

SARSOP.PolicyGraphGeneratorType
PolicyGraphGenerator
  • fast = false use fast (but very picky) alternate parser for .pomdp files
  • graph_max_depth maximum horizon of the generated policy graph. There is no limit by default.
  • graph_max_branch maximum number of branches of the generated policy graph. Shown will be top in probability. There is no limit by default.
  • graph_min_prob minimum probability threshold for a branch to be shown in the policy graph defaults to zero
  • graph_filename output name for the DOT file to be generated
  • pomdp_filename pomdp file input
  • policy_filename policy file input
source
SARSOP.generate_graphFunction
generate_graph(graphgen::PolicyGraphGenerator)

Generate a policy graph, see PolicyGraphGenerator to see the available options.

source