Skip to main content
. 2021 Apr 29;23(5):551. doi: 10.3390/e23050551
Algorithm 1 EM algorithm for DEC-POMDP
  • k0, Initialize θk.

  • Tmax(log(1γ)ε)/logγ1

  • while θk or J(θk) do not converge do

  •    Calculate p(x,z|x,z;θk) by Equation (20).

  •    //—E step—//

  •     α0(x,z;θk)p0(x,z;νk)

  •     β0(x,z;θk)r¯(x,z;πk)

  •    for  t=1,2,,Tmax  do

  •        αt(x,z;θk)x,zp(x,z|x,z;θk)αt1(x,z;θk)

  •        βt(x,z;θk)x,zβt1(x,z;θk)p(x,z|x,z;θk)

  •    end for

  •     F(x,z;θk)t=0Tmaxγtαt(x,z;θk)

  •     V(x,z;θk)t=0Tmaxγtβt(x,z;θk)

  •    //—M step—//

  •    Update θk to θk+1 by Equations (9)–(11).

  •     kk+1

  • end while

  • return  θk