Now we're going to look at NetLogo implementation of the Robbie the robot genetic algorithm This interface is a little bit complicated so I go through it So there is two parts to this interface, there's the genetic algorithm part, and then there's this us part that simulates Robbie in its environment. So let's run the genetic algorithm. we set the population size here I'm actually gonna set it to 60. One disadvantage of this net logo versions is kinda slow so I can't run it on the same size population that I ran the C language version and speed up this to make its fast as possible as to speed up all the way Okay so the population size 60. The mutation rate remember after parents are crossed over to create an offspring the offspring is mutated. this is the probability mutation at each value in the string there's a 0.05 probability that each value will be mutated this means on average about 12 mutations per string. It's kind of high but let's leave it there for now will experiment with this later on okay so now I do setup just takes a little while because slow and what you gonna see each the individuals in the population is represented here. I have 60 individuals in the population this is the first-generation where everything's random these each represent a random strategy and the vertical axis here represents the fitness see concede that these are all quite low fitness this one here is the highest fitness in the population the horizontal axis represents how similar these other individuals are to the best individual in terms of this stream that represents the strategy how similar the strings are okay so week either just run this forever we can run it for certain number generations and I've set the number generation to 50 so let's run it and over here it tells us the best fitness at each generation and also gives a representation of the strategy itself in terms of with the actions are in that strategy that's not that useful right now because a we can't really interpret how good a strategy is by looking at that string of course but now I can just watch this the population change overtime and you can see a plot of the best fitness of the best individual in each generation here you can see it's quite slow okay well got a generation 50 and it went from -74 of the best as the best fitness all the way to 7.5 is the best fitness. So now we can do is we can look at what's the best individual does in a given environment. So I set up the environment now that display switches to the individual his the best individual and here's what it's going to do in this particular configuration of cans for each of the cans the it to these orange dots is a can, so I can do is step through the best strategy well here with it set is at each time step this individual is just staying put okay it's ok it's zero fitness let's try a different environment here's a different configuration of can still just stay put here's a different configuration kids well there's some configuration of cancer this individual actually does something cuz it got higher than 0 average fitness but I guess we haven't found it yet ok it picked up a can it had a can on its site and picked up a can okay so it gets a score of 10 on this round alright well let some toggle display that goes back to our genetic algorithm display and lets try going for another 50 generations starting from the population we have now. I accidentally press Setup so I destroyed the population we had now so let's try that again go in generations so this is a different run it starts out with the different initial population have random strings so we'll see it different progression and now let me skip ahead to generation fifty in the video okay here we are. generation 50 this is done a little bit better the best fitness is 18 .5 so let's watch the best individual so we see stepping through it actually made some moves but it didn't pick up any cans. let's try different environment hear it actually picked up a can Ah it's picking up some cans so but that is going backwards and going again out stack anyway this is just to show you this model you can download it on from our course materials pages called RobbieGA.nlogo and you could play with it we're gonna do some exercises using it for the homework in this unit