Seite - 131 - in Adaptive and Intelligent Temperature Control of Microwave Heating Systems with Multiple Sources

4.2. IntelligentControl DefinethestatespaceS andtheactionspaceU; InitializeQ(s,u)arbitrarily foralls∈S,u∈U; Initializee(s,u) = 0 foralls∈S,u∈U; Initialize theadaptivecontrollerCad(Y,Ytar) (eitherMPCorNNC); Take thefirst controlactionU(k)at timek= 1basedontheequation U(k) =Cad(Y,Ytar) Ateachtimek (k>1) : repeat 1 Select theactivecontrollerCact(k)accordingto Cact(k)⇐      Cad(Y,Ytar), when Ymax<=Ytar, NoPower, when Ymin>=Ytar, Q(s,u), otherwise. ifCact(k) =Q(s,u) then 2 Calculate thecontrol actionU(k)andupdate theQvalues accordingto theQ(λ) learningalgorithm(seefigure 4.12); else 2 Calculate thecontrol actionU(k) =Cact(k); 3 Reset theeligibility tracematrixe(s,u) = 0 forall s∈S,u∈U; until the endof the controlprocess; Figure4.14. Procedures in the hybrid multi-agentQ(λ) learning control system. 131