MultiAgentDecisionProcess
AgentMDP Class Reference

AgentMDP represents an agent which uses a MDP-based policy. More...

#include <AgentMDP.h>

Inheritance diagram for AgentMDP:
[legend]

Public Member Functions

Index Act (Index sI, Index joI, double reward)
 Return an individual action based on state, last joint observation and reward. More...
 
 AgentMDP (const PlanningUnitDecPOMDPDiscrete *pu, Index id, const QTable &Q)
 (default) Constructor More...
 
 AgentMDP (const AgentMDP &a)
 Copy constructor. More...
 
void ResetEpisode ()
 Will be called before an episode, to reinitialize the agent. More...
 
 ~AgentMDP ()
 Destructor. More...
 
- Public Member Functions inherited from AgentFullyObservable
 AgentFullyObservable (const PlanningUnitDecPOMDPDiscrete *pu, Index id)
 (default) Constructor More...
 
 AgentFullyObservable (const AgentFullyObservable &a)
 Copy constructor. More...
 
 ~AgentFullyObservable ()
 Destructor. More...
 
- Public Member Functions inherited from AgentDecPOMDPDiscrete
 AgentDecPOMDPDiscrete (const PlanningUnitDecPOMDPDiscrete *pu, Index id)
 (default) Constructor More...
 
 AgentDecPOMDPDiscrete (const AgentDecPOMDPDiscrete &a)
 Copy constructor. More...
 
const
PlanningUnitDecPOMDPDiscrete
GetPU () const
 
- Public Member Functions inherited from SimulationAgent
virtual Index GetIndex () const
 Retrieves the index of this agent. More...
 
virtual bool GetVerbose () const
 If true, the agent will report more. More...
 
void Print () const
 Print out some information about this agent. More...
 
virtual void SetIndex (Index id)
 Sets the index of this agent. More...
 
virtual void SetVerbose (bool verbose)
 Set whether this agent should be verbose. More...
 
 SimulationAgent (Index id, bool verbose=false)
 (default) Constructor More...
 
virtual std::string SoftPrint () const
 Return some information about this agent. More...
 
virtual ~SimulationAgent ()
 Destructor. More...
 

Private Attributes

QTable _m_Q
 
size_t _m_t
 

Detailed Description

AgentMDP represents an agent which uses a MDP-based policy.

In particular, AgentMDP takes actions based upon states by looking up the maximizing action in the QTable (_m_Q) that this agent receives upon construction. This is in contrast to other AgentFullyObservable types that do some actual learning or planning during the simulation.

Constructor & Destructor Documentation

AgentMDP::AgentMDP ( const PlanningUnitDecPOMDPDiscrete pu,
Index  id,
const QTable Q 
)

(default) Constructor

AgentMDP::AgentMDP ( const AgentMDP a)

Copy constructor.

AgentMDP::~AgentMDP ( )

Destructor.

Member Function Documentation

Index AgentMDP::Act ( Index  sI,
Index  joI,
double  reward 
)
virtual
void AgentMDP::ResetEpisode ( )
virtual

Will be called before an episode, to reinitialize the agent.

Implements SimulationAgent.

References _m_t.

Member Data Documentation

QTable AgentMDP::_m_Q
private

Referenced by Act().

size_t AgentMDP::_m_t
private

Referenced by Act(), and ResetEpisode().