Unknown Speaker 0:00 Hi there, and welcome to the second part of this video series. My name is Julian Hird, and I'm a PhD student from the University of Bristol. I'm also part of a research group which focuses on swarming across scales, from another wo rld cancer treatments to the macro world of swarm robotics, which we're going to focus on in this video. As with the previous video, are two exercises which go alongside this session. The pre work, which was designed to be done before watching this video g ets you to look at the current state of the art for robotics, and will help you understand how swarm robotics is different. If you'd like to do this exercise, you can pause the video now. Before we begin, let's go over the aims of this video. Hopefully, by the end of it, you'll know what swarm robotics is and how it is different from other areas of robotics. We'll also touch on what a swarm robot consists of and how they come together to form a swarm system. Finally, we'll cover the advantages of peace syst ems and the current challenges associated with them. Okay, let's get to the fun bit robots. In the pre work, you will have seen a large variety of robot Books and pick one to research in more detail, they should have given you an idea of what robots can do and where they are used. And what should become immediately apparent is that robots and robotics is a really broad area, from robotic arms to flying robots and even soft robots. robots do many different jobs, and we travel in many different environments. You may have also noticed that things we take for granted are incredibly difficult to get robots to do reliably. You may have also seen but a lot of the robots featured in the suggested list are either remote controlled or operated fairly simple, controlle d environments. While these robots are still doing useful things, it does mean that they are restricted in what we can do with them and where we can use them. So researchers are always looking for ways to build more robust robotic systems that can cope wit h complex situations automatically. This brings us to swarm robotics. When you looked at swarm intelligence in the last video, you will have seen how swarms in nature are very robust systems. As they complete complex tasks and challenging environments, the idea with swarm robotics is to build a robotic system like a swarm, and hope to get some of these abilities. So with swarm robotics, we use lots and lots of simpler robots following simple rules, and then let the collective behavior that emerges, complete the tasks that we want to do. This is in contrast to other areas of robotics, which would generally train us a few or a single, larger, more complex machine to complete the task. So what does the swarm robotics system look like? So as mentioned before, a swarm robotic system will use lots of robots, we're talking large numbers of robots, hundreds, maybe even thousands. Each robot itself is quite simple. So, so simple, but they wouldn't be able to complete the task on their own. And each robot is also follo wing a simple set of rules. And this is determined by a control algorithm. control algorithm will look at what's happening around the robot and its local environment, what the neighboring robots are beside it, and produce local interactions, these local in teractions and in turn generate the emergent behavior, which will do the thing that we desire the swarm to do. So we can really think about individual robots in the swarm as having three core elements by his local control, which basically means the control the onboard, so the robot is making its own decision is following its control algorithm. And then these decisions are based on the local interactions and visa things are happening nearby. So this could be the robot sensing things, such as a light level ch anging around it. And this is local sensing. And this is sensing what's happening nearby. But it could also be communicating with nearby robots. And so we call that local communication. And then to form a swarm as a whole. You could have every robot in thi s world Being the same, and this is a homogeneous form, or you could have a heterogeneous swarm, where you've got maybe one or two or three different types of robot. Unknown Speaker 4:09 So why bother doing all this, that seems a very roundabout way of getting a robot to do something rather than just programming it to tell it what to do. Why rely on these local interactions leading to emergent behavior? Well, if we do build a robotic system like this, it does lead to a couple of advantages. Well, bec ause the swarm is comprised of many very similar robots, and this makes the robots interchangeable. In engineering, when we have lots of the same component in the system, we call this redundancy. And in swarms, we have lots of redundancy because we have lo ts of similar robots which interact with each other in a similar way. Essentially, lots of redundancy means that even if a single component in your system, in this case a robot breaks, then there's others that can step in to take his place. And this makes your system By robust. Also, the way that swarms interact on a local level means there's no single point of failure. So even if a, if a single communication link goes down, the swarm as a whole can keep operating. Another advantage of the way swarms work i s scalability, which means that the swarm can change in size really easily. This means you can add more robots if your task scales would be true balls. For example, if you had a swarm of robots who work together to push boxes around a warehouse, and sudden ly a new larger box arrives, well, you can add more robots to push this larger box. Whereas with a single robot solution, you'd have to design a whole new robot which will be able to carry this larger box. Hopefully, you should see those are very desirable features of a robotic system. And so how do we actually build one of these in reality? Well, we need to build lots and lots of swarm robots. A swarm robot is made up of the things we mentioned before local control, local sensors, local communication But t o be a robot, he needs a way of also moving around a locomotion system, and also a source of power to provide the robots with energy, and this usually is electrical power, so it will be a battery. All these elements can be found on the kilowatt, one of the most popular swarm robotic platforms, which costs around $20. To make. It uses infrared communication to talk to nearby killer bots by bouncing infrared light off the floor. This gives the robot local communication. This infrared LED can also send a dista nce to nearby robots by using the amount of light it receives. The visible light sensor also gives a reading of visible light shining down on the robot fees form the robots local sensing. The local controller is the microcontroller on the bottom of the rob ot. This is essentially a compact dedicated computer, similar to what you'd find in an Arduino. The robot moves using two vibrational motors allowing it to kind of scuttle along a flat surface. By controlling the speeds of these motors. The robot can move forwards or turn In a slight circle, and this forms a very basic locomotion system. Finally, the whole robot is powered through a rechargeable battery. Now, if we put loads of these robots together, they can do interesting things. For example, in this vide o 1000 killer robots are doing shape formation. They basically organize themselves into any shape we desire, and that we're doing this all with simple rules. a different type of swarm is Futterman swarm, which operates in a similar way to that of termites. When building their nests. Each robot climbs onto the structure carrying a block. And it's following a simple set of rules about where to place a block based on the block site has seen has already been placed. different rules produce different structures. And a computer program is used to generate the set of rules for a particular structure. Unknown Speaker 7:51 The system is very robust as you can remove blocks and then the next robot will recognize the gap because the gap is what caused the block to be placed in First place. This is in contrast to telling each robot exactly where to place their block. And so if a single block was out of position, the whole structure could collapse. So we've seen how swarms are built. We know why very exciting, and we've seen a couple of examples of what they can do. But swarm robotics also has some challenges before it can be used widely. As a swarm intelligence designing the local rules say about the robots do what we wish and also importantly, don't do anything we don't want them to do is difficult. Another challenge is the use of large numbers of robots requires about the individual robots are small and low cost, which can limit their complexity. But the price of components is always coming down as robotic compone nts start being mass produced, and new algorithms are always being developed, but require less and less sophistication. There's also the question of how humans will interact with swarms and where swarms make sense to be used in the real world. And that bri ngs us to the end of this video on swarm robotics. I hope you have enjoyed it. If you'd like to learn more, there will be resources shown over examples as well. robotics in the resource document associated with this video series. Here you can also find pos t video exercise for you to try. As mentioned, swarms are now being designed with real world applications in mind. This exercise will get you to design your own swarm for a task of your choosing. This could be finding survivors after an earthquake or maybe cleaning up an oil spill. Thanks for listening Transcribed by https://otter.ai