Reinforcement Learning Toolbox 2.0 last updated:

General Documentation Manual Tutorial Class Reference Master Thesis Examples Related Papers Downloads Links News mailto:webmaster Main Page     Class Hierarchy   Compound List   File List   Compound Members   File Members CSemiMarkovDecisionProcess Class Reference Class for providing the general Functions for the learning Environment. More... #include <cagent.h> Inheritance diagram for CSemiMarkovDecisionProcess: List of all members. Public Member Functions   CSemiMarkovDecisionProcess ()   ~CSemiMarkovDecisionProcess () virtual void  sendNextStep (CStateCollection *lastState, CAction *action, CStateCollection *currentState)   Sends the next Step to all Listeners. I.e that if the Action is a finished MultiStepAction. CAction *  getLastAction ()   Returns the last Action sent to all Listeners. virtual void  startNewEpisode ()   Sends to all Listeners that a new Episode occured and updates currentEpisodeNummer. int  getCurrentEpisodeNumber ()   Returns the number of Episodes. int  getCurrentStep ()   Returns the number of steps. int  getTotalSteps () virtual void  addAction (CAction *action)   Adds an Action to the ActionSet of the SMDP. virtual void  addActions (CActionSet *action) Protected Attributes CAction *  lastAction int  currentEpisodeNumber int  currentSteps int  totalSteps bool  isFirstStep Detailed Description Class for providing the general Functions for the learning Environment. CSemiMarkovDecisionProcess is the super class of all "acting" agents. It maintains a list of available Actions for the SMDP. It also loggs the number of Episodes and the Number of Steps done in the current Epsiode. It provides the functionality for sending a Semi-Markov Step, but its only able to send PrimitiveActions. For extended Actions you have to use the hierarchicalMDP. /par The class is a subclass of CDeterministicController in order to make any Controller assigned to the SMDP a deterministic Controller. The CDeterministicController Object is always the first Object to be informed about the new Step, and its not in the ListenerList (Recursions). This feature is needed for learning the exact Policy of the agent (see CSarsaLearner), so the agent can be used as estimation policy. See also: CAgent CHierarchicalalSemiMarkovDecisionProcess Constructor & Destructor Documentation CSemiMarkovDecisionProcess::CSemiMarkovDecisionProcess (   )  CSemiMarkovDecisionProcess::~CSemiMarkovDecisionProcess (   )  Member Function Documentation virtual void CSemiMarkovDecisionProcess::addAction (  CAction *  action  )  [virtual]   Adds an Action to the ActionSet of the SMDP. virtual void CSemiMarkovDecisionProcess::addActions (  CActionSet *  action  )  [virtual] int CSemiMarkovDecisionProcess::getCurrentEpisodeNumber (   )  [inline]   Returns the number of Episodes. int CSemiMarkovDecisionProcess::getCurrentStep (   )  [inline]   Returns the number of steps. CAction* CSemiMarkovDecisionProcess::getLastAction (   )    Returns the last Action sent to all Listeners. int CSemiMarkovDecisionProcess::getTotalSteps (   )  [inline] virtual void CSemiMarkovDecisionProcess::sendNextStep (  CStateCollection *  lastState, CAction *  action, CStateCollection *  currentState )  [virtual]   Sends the next Step to all Listeners. I.e that if the Action is a finished MultiStepAction. Reimplemented from CSemiMDPSender. virtual void CSemiMarkovDecisionProcess::startNewEpisode (   )  [virtual]   Sends to all Listeners that a new Episode occured and updates currentEpisodeNummer. Reimplemented from CSemiMDPSender. Reimplemented in CAgent, and CHiearchicalAgent. Member Data Documentation int CSemiMarkovDecisionProcess::currentEpisodeNumber [protected] int CSemiMarkovDecisionProcess::currentSteps [protected] bool CSemiMarkovDecisionProcess::isFirstStep [protected] CAction* CSemiMarkovDecisionProcess::lastAction [protected] int CSemiMarkovDecisionProcess::totalSteps [protected] The documentation for this class was generated from the following file: cagent.h