func New(cfg Config, mdp discrete.MDP) (this *Agent) {
this = new(Agent)
this.cfg = cfg
this.mdp = mdp
this.mdpo = discrete.NewMDPOracle(this.mdp, 0)
this.s = fsss.New()
this.s.Cfg = this.cfg.FS3
this.s.NumActions = this.mdp.NumActions()
this.s.Gamma = mdp.GetGamma()
this.s.Vmin = this.mdp.GetTask().Reward.Min / (1 - this.s.Gamma)
this.s.Vmax = this.mdp.GetTask().Reward.Max / (1 - this.s.Gamma)
this.stepsWithPlanner = 0
return
}
func ValueIteration(qt *discrete.QTable, mdp discrete.MDP, epsilon float64) (numIterations int) {
//fmt.Fprintf(os.Stderr, "+ValueIteration\n")
//fmt.Println(mdp.GetGamma())
//defer fmt.Fprintf(os.Stderr, "-ValueIteration\n")
var error float64
for {
numIterations += 1
//fmt.Printf("iteration %d\n", numIterations)
error = 0
for s := range mdp.S64() {
for a := range mdp.A64() {
saError := BackupStateAction(qt, mdp, s, a)
error = math.Fmax(error, saError)
}
}
//fmt.Printf("QT\n%v\n", qt)
//fmt.Fprintf(os.Stderr, "error %f\n%v\n", error, qt)
if error < epsilon {
return
}
}
return
}
func BackupStateAction(qt *discrete.QTable, mdp discrete.MDP, s discrete.State, a discrete.Action) (error float64) {
var nq float64
for n := range mdp.S64() {
ev := mdp.T(s, a, n)
ev *= qt.V(n)
nq += ev
}
nq *= mdp.GetGamma()
nq += mdp.R(s, a)
error = math.Fabs(nq - qt.Q(s, a))
qt.SetQ(s, a, nq)
return
}