AIToolbox::POMDP::RTBSS< M > Class Template Reference

This class represents the RTBSS online planner. More...

#include <AIToolbox/POMDP/Algorithms/RTBSS.hpp>

Public Member Functions
	RTBSS (const M &m, double maxR)
	Basic constructor. More...
std::tuple< size_t, double >	sampleAction (const Belief &b, unsigned horizon)
	This function computes the best value for a given belief and its value. More...
const M &	getModel () const
	This function returns the POMDP model being used. More...

Detailed Description

template<IsModel M>
class AIToolbox::POMDP::RTBSS< M >

This class represents the RTBSS online planner.

This algorithm is an online planner for POMDPs. It works by pretty much solving the whole POMDP in a straightforward manner, but just for the belief it is currently in, and the horizon specified.

Additionally, it uses an heuristic function in order to prune branches which cannot possibly help in determining which action is the actual best. Currently this heuristic is very crude, as it requires the user to manually input a maximum possible reward, and using it as an upper bound.

Additionally, in theory one would want to explore branches from the most promising to the least promising, to maximize pruning. This is currently not done here, since an heuristic is intrinsically determined by a particular problem. At the same time, it is easy to add one, as the code specifies where one should be inserted.

This method is able to return not only the best available action, but also the (in theory) true value of that action in the current belief. Note that values computed in different methods may differ due to floating point approximation errors.

Constructor & Destructor Documentation

RTBSS()

template<IsModel M>

AIToolbox::POMDP::RTBSS< M >::RTBSS	(	const M &	m,
		double	maxR
	)

Basic constructor.

Parameters

m	The POMDP model that POMCP will operate upon.
maxR	The max reward obtainable in the model. This is used for the pruning heuristic.

Member Function Documentation

getModel()

template<IsModel M>

const M & AIToolbox::POMDP::RTBSS< M >::getModel

This function returns the POMDP model being used.

Returns

The POMDP model.

sampleAction()

template<IsModel M>

std::tuple< size_t, double > AIToolbox::POMDP::RTBSS< M >::sampleAction	(	const Belief &	b,
		unsigned	horizon
	)

This function computes the best value for a given belief and its value.

Parameters

b	The initial belief for the environment.
horizon	The horizon to plan for.

Returns

The best action and its value in the model.

---

The documentation for this class was generated from the following file:

• include/AIToolbox/POMDP/Algorithms/RTBSS.hpp