 Bonjour tout le monde, merci d'inviter moi à montrer ma recueillie, je suis Bruno Bastien, je suis currently working at the Max Planck for Intelligent Constance in the department of Ian Cuisine, et je vais vous montrer quelques résultats en anglais et en avec d'eux, qui ont été master students in the lab. Je vais vous parler d'un progrès pour les hommes et les méchants, par méchants, je veux dire méchants, comme dans le film. Pour les personnes who haven't seen the movie, it's a 1999, American movie, where people are living together in a virtual world. So usually it's quite common to use matrix as a short end for virtual reality. When people say how it's matrix, it means it's virtual reality. So by virtual reality, what we mean, is that it's an artificial environment which is experienced with sensory stimuli, provided by a computer and it switch one action partially determined what happened in the movie, partially determined what happened in the environment. So what that mean is that like, your perceptual environment is going to be immersed in a numerical simulation and it's going to be interactive. The action you are going to do are going to modify the virtual world. So what you might have in mind is the headset, which are like Oculus Rift and this type of stuff. So the first one is from 1968. So what's happening is headset, we're just in front of this Juviel field, and it's going to allow him here to see image in the direct environment, actually. So you can see like he's seeing a cube. So it's 1968, so no, we are doing that much better. The basis of all this technique of virtual reality is the anamorphosis, which is an illusion based on perspective. So there is this artist called Feliceva Rimi, who is painting this structure on the environment. So when you look then from the white place, you are seeing this grid that appears in space. But if you are at a different place, then the perspective is not correct and you can't understand the image anymore. The idea of virtual reality is then to account in real time for the position of the viewer and to modify the perspective so that the image appears correct for the viewer. So this example from 2007, so here they use like a remote, so it's a Nintendo remote control. And basically if you look at the screen, what you see is that like the image looks static and if you move nothing up here. But at some point it's going to modify system so that like the position of the camera is going to be tracked in real time. And the image are going to be changed for account for the position. And so now we can give the illusion that there is like object in the truly space. So it's important to understand at this point that like this illusion only works for one perspective. So for each user, we need a different system. So now the question is that like we are interested in animals. So the headset is not going to work, but this type of technique can be used for animals. So recently, so John Stowe, Max of Bauer and Andrew Stroh in Vienna started to develop system for freely moving animals and here it's a fly. So you have a fly here, you have like multiple camera tracking the animal and you have a projection here that is like this virtual environment. So the idea is to give the illusion to the fly there is a pilon in the middle of the room. So what you see is that like the fly is going to move, the perspective is modified at each point so that to account for the perspective of the fly here. And if you look at all the trajectory, the fly is going to avoid the object. So it is no possible to give the illusion to individual that there is object in their environment like virtual object. So I'm going to use a more restrictive definition of matrix. In the case here what I'm going to use is the fact that like it's an immersive and interactive virtual environment that is shared. Meaning like we're going to have like multiple animal in the same virtual environment and where such an interaction can happen. So there are multiple motivations to try to reach this kind of system but I'm going to give you one that is more like on the theoretical side let's say. In collective behavior, if you look at the classical model of collective behavior they are all going to have like some implicit of distance. If you look by example the V-check model what you say is that like the animal is going to interact with this closest neighbor. If you look now by example at the Cousine model you are going to say like in the certain zones they are going to be a repel, in certain zones they are going to be a line etc. But all this model are going to have this type of implicit like the animal have directly access to the distance of this neighbor. But they don't have directly access to this information. This information needs to be integrated through their perception, in particular through their visual field. So if we have access to a virtual reality system then the goal is to control and to perturb individually the visual field of each individual. So it's a, sorry for that movie, so it go with like some research by example by Ariana, some Tricaine cuisine lab where they started to look the influence of the visual field on the organization of this one. But there is really a need, like if we want to go further we need to find a way to get closer to the visual field. So what is the relation between the visual field and the movement and can you control and perturb the visual field of each individual. Then virtual reality. An important point for us is that we want to be able to generate social interaction in a virtual environment. So how do we generate social interaction in a virtual environment? So if I just look like the definition of social interaction social interaction is an interaction between two or more individual. So, fine. But the question now is that like if I have the individual interacting through their virtual orientation do I still get a social interaction? So how can I come for that? So that's what I'm going to discuss mainly here. So the system we have is called the fichvia developed by Loup-Baillot. So we have a ball here that you can see here on which we are going to project images. Inside the box we have like four cameras that are going to track in real time the position of the individual. So by acting like for the position at each time we are going to modify the perspective and we can create an interactive virtual environment and give the illusion of the animal that there is object in the virtual space. So the first thing we need to do before is to try to get a feeling of what is the social interaction between two fiches. So here I'm only going to look at zero fiches larvae. So if you put two fiches together so that's just that's the symbol in the real world but it's just a truly foundation. So what you see is that like they are going to interact together. So one fish is going to follow the other and they are going to interact more or less in the same place. And sometimes this interaction is going to stop again. So that's the basis of the social interaction between two individuals. So all the analysis I'm going to do on this presentation is going to be based on this simple analysis. Meaning like I'm going to look at my focal fish and in the referential of the focal fish I'm going to look the position of the other fish. If I do that multiple times I can create a density map that is going to tell me where I can find like the other fish. So if I do that now for my two fish where I see that like in the case in my control case here I just have two normal fish. I'm going to have like this map. So it's ever going to be my fish is here, the other fish is ever in front or in the back. Meaning I'm going to create like a pair of leader followers like ever the fish is going to be followed and the other one is going to follow it or it's going to lead. What we try to do is also we knocked out the lateral line so the fish are the lateral line on the side that allows them to sense hydrodynamic flow that is thought a lot to be central to collective behavior in fish. If you knock out like the lateral line you still get the same pattern. The two fish still get like a similar interaction even in absence of perception of the hydrodynamic flow. You have like one that is going to be in front what is going to be in the back. So in that case I get like both on the same graph because currently we can't conserve identity during the analysis. So now what we do is that like we look at the movement of a fish. So that's where the fish larvae which has like this burst and glide movement with like some oscillation during the burst and we make a virtual presentation of the annual. So that's our virtual zebra fish that we're going to put in the virtual world. So now from the point of view of the fish. So that's the projection. So that's the virtual fish and the real fish is here. So here we will fish out of the illusion that the virtual fish is somewhere in the middle. So here it's going to follow it on. So that's the first thing. But now what we have is that like we have multiple systems in the department. We have like five systems. So that's the fish match weeks if you want from the outside. So we have multiple systems together and give the illusion they are in the same virtual environment. Namely like in one ball I'm going to have this real fish looking at this virtual fish which is in fact controlled by this other real fish and the other ball which is looking at this virtual fish which is in fact this one. So they are interacting in two different ball but in the same virtual world. So if I put my two fish together I'm going to have like this thing. So the two fish here are not in the same virtual world. But I'm going to see similar dynamic from what we are seeing together. One fish is going to follow the other and they are interacting more or less than something. And sometimes the interaction like break and come back. So it seems that we are able to create like some things that look like a social interaction. More in detail like if we look like if we look like, so that's the first fish, that's the second fish no the density map is going to be more clear. One fish has the other fish in front is going to follow it. And the other fish is going to have the fish in the back. It's going to be the leader in that case. This is really variable. If we do the experiment multiple times we create this pair of leader follower almost all the time. So now if I can put two animals and get, I can also put like four of them. So just to show you what's happened if you put like four individuals each one of them is in a different ball but they are interacting together and they are going to school in the virtual world. But we can go even further like know that we know like two individuals seems to be in the same way that in the real world we can also remove the feedback. We don't need to have like the connection between the two individuals. I can just preprogrammatrack and see if my fish is still trick. So here I have like that is following a virtual fish that is going on track and what we see is that like this fish is going to follow the virtual fish in a way that seems to mimic the social interaction we are seeing in the real world and in the matrix. But if we are able to do that know we can like try to map precisely and accurately the parameter space. So for example what we did is that we use this circle here and we did it for different speed. Here a virtual fish we put it at certain depth so that it can be visible and then I'm looking what's happening when I change the speed of my virtual fish. So here it's for low speed so that's the internal distance between the real fish and the virtual fish. What I see is that like when the speed is small or comparable to the speed of the real fish they are going to be close together and you can see here when you change the speed then this interaction is going to disappear. So this interaction that we see also happen only in cases where the range is similar to the range of the real fish. But no we can also say that maybe they are just following like anything we put in the bowl but we can change the virtual representation. So we had something that was like very realistic that looked like a fish that seems to behave like a fish and so if you look now like for different and you look at the internal distance you see like in the case you have a real fish it's going to be close by most of the time it's going to be like less than 7 cm but if you look now what's happening if you just put a sphere then you don't have interaction anymore so in a way the interaction is selective so based on all this criteria we can more or less determine that there exists like social interaction inside the virtual world the interaction is independent of the lateral line so it's independent of like the perception of the dynamic flow it reproduise the feature of the social interaction in the real world like leader follower distant of interaction it's similar to interaction without feedback so the feedback doesn't seem to be even like a feature necessary for a social interaction it's observed in a large range of parameters that is realistic so with that force speed with that like for a lot of different parameters and it's also selective on the visual appearance it doesn't work with a sphere I'm just going to spend like one minute on perspective so where do we go knows that we have that so one part of the problem is trying to map like the whole parameter space we can modify and tune all the parameters we want so for example if we use scale so that's like the proper size of the fish that's like 10 times the size of the fish that's like a tiny fish we don't have much interaction when the fish is too small but when it reach the right size and when it starts to grow again this distance increase if I put no my two fish in the matrix and change their size I'm going to see that's like they're going to be close together and I'm going to increase the size so we have no way to modify directly the interaction between the two I can like tune like the relation and the distance at which they want to be together the work that we are working on is that we can modify like the new topology so by example I can have like this fish that like see this one that is seen by this one but it's seen like some virtual fish and have like some really complex topology and try to see how like information is transferred inside like this complex new topology adjust one side and so the last slide is this one so that experiment by Liangli which is like a post doc in the lab at this point and what he tried to do is that he just put two fish together two virtual fish going straight at certain distance and what he observes is that like if we look at the position at which the fish is when this inter-individual distance between the two virtual fish is small the two fish, the real fish is in the middle so it seems to make consensus between both positions but some distance is going to bifurcate and going to choose between one fish or the other on a this type of tool we also have a way to try to go like further into like decision making process of the individual finally because I still have one minute one of the goal like for the future is Simon Jean Jean is a field biologist in the lab make like truly scan of coral for his experiment and one of the goal also is like in the future is to reproduce natural environment and try to do like field experiment directly in the lab thank you