write_POMDP function

Read and write a POMDP Model to a File in POMDP Format

Reads and write a POMDP file suitable for the pomdp-solve program.

write_POMDP(x, file, digits = 7, labels = FALSE)

read_POMDP(file, parse = TRUE, normalize = FALSE, verbose = FALSE)

Arguments

• x: an object of class POMDP .
• file: a file name. read_POMDP() also accepts connections including URLs.
• digits: precision for writing numbers (digits after the decimal point).
• labels: logical; write original labels or use index numbers? Labels are restricted to [a-zA-Z0-9_-] and the first character has to be a letter.
• parse: logical; try to parse the model maotrices. Solvers still work with unparsed matrices, but helpers for simulation are not available.
• normalize: logical; should the description be normalized for faster access (see normalize_POMDP())?
• verbose: logical; report parsed lines. This is useful for debugging a POMDP file.

Returns

read_POMDP() returns a POMDP object.

Details

POMDP objects read from a POMDP file have an extra element called problem which contains the original POMDP specification. The original specification is directly used by external solvers. In addition, the file is parsed using an experimental POMDP file parser. The parsed information can be used with auxiliary functions in this package that use fields like the transition matrix, the observation matrix and the reward structure.

The range of useful rewards is restricted by the solver. Here the values are restricted to the range [-1e10, 1e10]. Unavailable actions have a reward of -Inf which is translated to -2 times the maximum absolute reward value used in the model.

Notes:

The parser for POMDP files is experimental. Please report problems here: https://github.com/mhahsler/pomdp/issues.

Examples

data(Tiger)

## show the POMDP file that would be written.
write_POMDP(Tiger, file = stdout())

References

POMDP solver website: https://www.pomdp.org

See Also

Other POMDP: MDP2POMDP, POMDP(), accessors, actions(), add_policy(), plot_belief_space(), projection(), reachable_and_absorbing, regret(), sample_belief_space(), simulate_POMDP(), solve_POMDP(), solve_SARSOP(), transition_graph(), update_belief(), value_function()

Author(s)

Hossein Kamalzadeh, Michael Hahsler