Expression::Expression(const std::wstring& s, NfaStates& states, TransitionSet& symbols) : m_ch(0)
{
m_LiteralString = s;
// m_Type = ExpressionTypeSingle;
m_Left = nullptr;
m_Right = nullptr;
//(start) -a-> () -b-> ....() -c-> (end)
//cria uma cadeia de "ands" com cada caracteres o start e end sao as extremidades
m_StartState = states.New();
m_EndState = 0;
NfaState* state = m_StartState;
for (size_t i = 0 ; i < s.size(); i++)
{
if (i == 0)
m_ch = Transition(s[i]);
NfaState* pNew = states.New();
Transition ch(s[i]);
symbols.insert(ch);
AddTransition(state, ch, pNew);
state = pNew;
}
m_EndState = state;
//AddTransition(state, m_EndState);
}
void MakeGetCharIndex(const TransitionSet& symbols, std::wostream& stream, bool code)
{
int index = 0;
TransitionSet::const_iterator it = symbols.begin();
if (code)
{
stream << L"int GetCharIndex(wchar_t ch)" << endl;
stream << L"{" << endl;
stream << L" switch (ch)" << endl;
stream << L" {" << endl;
}
else
{
stream << L"//Symbols" << endl;
}
for (; it != symbols.end(); ++it)
{
if (it->m_ch != EmptyTransition)
{
if (code)
{
stream << L" case " << ToCharRepresentation(it->m_ch) << L": return " << index << L";" << endl;
}
else
{
stream << ToCharRepresentation(it->m_ch) << endl;
}
}
index++;
}
if (code)
{
stream << L" }" << endl;
stream << L"}" << endl;
}
}
void State::insertTransitions(const TransitionSet& transitions)
{
outTransitions.insert(transitions.begin(), transitions.end());
}
Expression::Expression(std::unique_ptr<Expression> A,
std::unique_ptr<Expression> B,
ExpressionType type,
NfaStates& states,
TransitionSet& symbols)
{
// m_Type = type;
m_Left = std::move(A);
m_Right = std::move(B);
if (type == ExpressionTypeAnd)
{
/// AB
/// |----------------| |-----------------|
/// | A | e | B |
/// | ( ) -----> ( )---------->( ) -----> (O) |
/// |----------------| |-----------------|
AddTransition(m_Left->m_EndState, m_Right->m_StartState);
m_StartState = m_Left->m_StartState;
m_EndState = m_Right->m_EndState;
}
else if (type == ExpressionTypeOr)
{
/// A|B
/// |----------------|
/// e | A | e
/// --------->( ) -----> ( )---------
/// | |----------------| |
/// | v
/// ( ) (O)
/// | ^
/// | |----------------| |
/// | | B | |
/// --------->( ) -----> ( )---------
/// e |----------------| e
m_StartState = states.New();
m_EndState = states.New();
AddTransition(m_StartState, m_Left->m_StartState);
AddTransition(m_StartState, m_Right->m_StartState);
AddTransition(m_Left->m_EndState, m_EndState);
AddTransition(m_Right->m_EndState, m_EndState);
}
else if (type == ExpressionTypeRange)
{
m_StartState = states.New();
m_EndState = states.New();
AddTransition(m_StartState, m_Left->m_StartState);
AddTransition(m_StartState, m_Right->m_StartState);
for (wchar_t ch = m_Left->m_ch.m_ch + 1 ; ch < m_Right->m_ch.m_ch; ch++)
{
symbols.insert(Transition(ch));
std::unique_ptr<Expression> pNew(new Expression(Transition(ch), states));
AddTransition(m_StartState, pNew->m_StartState);
AddTransition(pNew->m_EndState, m_EndState);
}
AddTransition(m_Left->m_EndState, m_EndState);
AddTransition(m_Right->m_EndState, m_EndState);
}
else
{
assert(false);
}
}
//=============================================================================
int main(int argc, char* argv[]) {
string userName = "******";
TabularQ q;
Param param;
ActionIterator a_iter;
auto greedy = rl::sa::greedy(a_iter);
auto argmax_critic = rl::sa::q_learning(q,greedy,param);
Simulator simulator;
ActionToss a_toss;
int episode,frame;
TransitionSet transition;
Reward sum = 0.0;
int episode_length;
auto explore_agent = rl::agent::epsilon_greedy(argmax_critic, a_toss, param);
auto test_agent = rl::agent::greedy(argmax_critic);
cout << "\n\t\t< Human-aware Decision Maker >" << endl;
cout << "> Please press ENTER to start..." << endl;
getchar();
/* Load the parnter data and set the paramter to Q-learning agent */
gsl_vector * dataLearned = gsl_vector_alloc (90);
FILE * fUserData = fopen ((userName + ".dat").c_str(), "r");
if (fUserData != NULL) {
cout << "\n> INFO: Loading user, " << userName << ", preference data...\n" << endl;
gsl_vector_fscanf (fUserData, dataLearned);
fclose (fUserData);
argmax_critic = dataLearned;
}
gsl_vector_free(dataLearned);
for(episode = 0, frame = 0; episode < NB_EPISODES; ++episode) {
std::cout << ">\t\t< Episode " << std::setw(6) << episode + 1 << "/" << NB_EPISODES << " >" << endl;
simulator.restart();
rl::episode::sa::run_and_learn(simulator, explore_agent, argmax_critic, transition, 0, episode_length);
for (auto iter = transition.begin(); iter != transition.end(); ++iter) {
sum += (*iter).reward(); // *iter is a transition
//cout << "> Action: " << (*iter).currentAction() << ", State_toa: " << (*iter).currentState().robotBelief << ", State_alr: " << (*iter).currentState().personAttention << endl;
}
cout << "> Total reward during episode : " << sum << std::endl;
/* Store the param into file */
const gsl_vector* parnterPreference = argmax_critic.parameter();
FILE * fparam = fopen((userName + ".dat").c_str(), "w");
gsl_vector_fprintf (fparam, parnterPreference, "%.5g");
fclose (fparam);
/* Store the reward into file */
FILE* freward = fopen((userName + "Reward.txt").c_str(), "a+");
fprintf(freward, "%f\n", sum);
fclose(freward);
/* Clear data for next episode */
sum = 0.0;
transition.clear();
transition.resize(0);
//getchar();
}
//try {
// simulator.restart();
// for (int k = 0; k < MAX_EPISODE_DURATION; ++k)
// simulator.timeStep(test_agent.policy(simulator.sense()));
//} catch(rl::exception::Terminal e)
//{ }
transition.clear();
transition.resize(0);
simulator.restart();
rl::episode::sa::run_and_collect(simulator, test_agent, transition, 0, episode_length);
// Let us read the collected transitions
for (auto iter = transition.begin(), iter_end = transition.end(); iter != transition.end(); ++iter) {
sum += (*iter).reward(); // *iter is a transition.
//cout << "> Action: " << (*iter).currentAction() << ", State_toa: " << (*iter).currentState().robotBelief << ", State_alr: " << (*iter).currentState().personAttention << endl;
}
cout << "Total reward during episode : " << sum << std::endl;
cout << endl;
cout << "Please press ENTER to left..." << endl;
getchar();
return 0;
}
