Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Optimization of simulated biological multi-agent systems by means of evolutionary processes Alexander Hörnlein Christoph Oechslein Frank Puppe Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Motivation / Problem • Optimization of behavior in respect of – explicit evaluation function – implicit evaluation function e.g. “the agents have to survive a certain period” • Calibration towards a predefined target behavior e.g. “the agents should act exactly as in real life” Optimization of simulated biological multi-agent systems by means of evolutionary processes 2 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Evolution as optimization • Population of potential solutions • Evaluation by means of “natural selection” • Iteration: Survivors (i.e. highly fit individuals) reproduce Optimization of simulated biological multi-agent systems by means of evolutionary processes 3 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Reproduction • Mutation – Offspring differs slightly - possibly advantageous – local search • Recombination – Child possibly unites the advantages of both parents – global search Optimization of simulated biological multi-agent systems by means of evolutionary processes 4 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Behavior in SeSAm Agent • Rules • Activities Activity1 IF (in activity1) AND Condition THEN activity3 • Parameters • Memory • Perception Optimization of simulated biological multi-agent systems by means of evolutionary processes Activity2 Activity3 Action1 Action2 ... 5 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg GP approach: Mutation operators activity • Change Add newnumeric activity terminals • Add Change newsymbolic rule terminals Parameter ab += 10 25 Flee from agent x Approach Increase speed Focus on earth • Change rule non-terminals • Delete activity action • Delete Add action rule Optimization of simulated biological multi-agent systems by means of evolutionary processes 6 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Advantage • Extremely powerful • Little constraint by initial structure of behavior Disadvantages • Development of unnecessary or unwanted complexity • Restrictions are difficult to define/set • Slow • Hard to implement within SeSAm Optimization of simulated biological multi-agent systems by means of evolutionary processes 7 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg GA/ES approach: Mutation operators activity Parameter a += 10 25 Approach agent x Increase speed • Change numeric terminals that’s it in principle. Optimization of simulated biological multi-agent systems by means of evolutionary processes 8 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Applicability of GA/ES approach within SeSAm • Actions – Use numerical terminals – Can be controlled by probabilities • Rules – Condition-parts use numerical terminals – Action-parts can be controlled by probabilities Optimization of simulated biological multi-agent systems by means of evolutionary processes 9 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Model modification • Define rules for any reasonable transient • Let evolution weight them • Treat actions accordingly Optimization of simulated biological multi-agent systems by means of evolutionary processes 10 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Advantages Disadvantage • Sufficient powerful • Not extremely powerful • Easy to restrict: Evolution can’t break boundaries of predefined behavior • Fast • Implementation within SeSAm is ‘straight-forward’ Optimization of simulated biological multi-agent systems by means of evolutionary processes 11 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg SeSAm genes RULE: IF ENERGY > gene0 THEN MOVE gene0: lower boundary upper boundary (initial) value (initial) standard deviation distribution dominance [ lower boundary Optimization of simulated biological multi-agent systems by means of evolutionary processes (initial) standard deviation (initial) value ] upper boundary 12 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg SeSAm genomes agent role behavior genome declaration genome egg storage gene0 declaration gene1 declaration ... allele0-0 allele0-1 ... allele1-0 allele1-1 ... ... family attribute gene0 gene1 Optimization of simulated biological multi-agent systems by means of evolutionary processes 13 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Polyploid genome • Treated threadwise • Treated genewise dominance mutation Optimization of simulated biological multi-agent systems by means of evolutionary processes dominance mutation 14 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Possibilities for the gene-expression • dominant/recessive • ‘intermediary’ value0 • weighted ) value (dominance 1 value expression value0 # alleles (dominance ) i value1 meta gene i i i i i i value2 Optimization of simulated biological multi-agent systems by means of evolutionary processes 15 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Application from individuals to colonies Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Insects’ behavior hunt set marker seek marker transport to nest prey from own reservoir fight seek new nest brood care from nest reservoir idle mate lay egg insects feed on brood grow Optimization of simulated biological multi-agent systems by means of evolutionary processes feed on nest reservoir feed 17 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Insects’ genes hunt-factor hunt brood care-factor prey transport to nest egg level genes from own reservoir fight from nest reservoir idle queen-factor mate feed on brood lay egg seek new nest brood care feed on nest reservoir grow feed Optimization of simulated biological multi-agent systems by means of evolutionary processes energy level genes 18 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Initial insects’ world Optimization of simulated biological multi-agent systems by means of evolutionary processes 19 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Insects’ world after 150,000 ticks Optimization of simulated biological multi-agent systems by means of evolutionary processes 20 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Changes of gene-pool queen-factor hunt-factor Optimization of simulated biological multi-agent systems by means of evolutionary processes brood care-factor 21 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg More changes of gene-pool initial egg energy energy portion ant Optimization of simulated biological multi-agent systems by means of evolutionary processes energy portion brood 22 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg Results & Discussion • Successful evaluation in three scenarios • ES/GA approach powerful and easy to use ? Use of explicit evaluation function for greater applicability ? Accelerate optimization (through parallelism) Optimization of simulated biological multi-agent systems by means of evolutionary processes 23 / 23 Lehrstuhl für Künstliche Intelligenz - Univ. Würzburg