In this video, we give some example of some student projects where students have worked in a robotic theme.
First is about robot learning. Now, we saw in an earlier video that robots can learn to move around by having a set of actions, and associated with each action is the probability of choosing it. And if the action is successful, the probability is increased. And initially, we needed five sets of actions which were chosen by us which made the problem perhaps a bit easy. So I asked Isaac Ashdown to investigate if a robot could automatically determine how many sets there should be. In order to do that and to demonstrate what he was doing, for each of the actions which had probabilities, they were depicted like this. Now, for this work, we had nine actions, not the four we had before.
So each motor could be forward, stopped or back. And these red marks indicate the probability of the action. And these will go up and down as the program runs. So at the moment, for instance, going forward has a high probability. Going back with the left motor, forward with the right, has a very low probability.
Isaac wrote a program to demonstrate his ideas. And we have here nine different sets of actions, and later on, we tried even more. And as the program runs, so the probabilities will change. So if I start running it, you’ll see we are out in the open so these probabilities are changing. Now, we’re going down here, around here, because with the robot can see something. So we’re in a different situation. As it moves around, different of these areas, the probabilities, changed, and the end, we started to merge areas which are quite close. So where we started off with nine actions, we could end up with fewer. Overall, it was a successful project.
Second project is about augmented reality, which is when we project on something real, something artificial. This was done by Chris Tingley who had a robot moving around an environment. And that robot had emotions such as a curiosity, C, and anger or A. And what we see there is these emotions being projected onto the scene. So you can see, this is very curious, and there’s hardly any anger. Now, the emotions were set by information from the senses. So for instance, if an object was further away, the robot became more curious. If it was stuck in a corner, it got angry. And then those emotions we used to set the speed of the robot.
For instance, if the robot became very curious, then it would start to go faster. Otherwise it would slow down. And as I say, these emotions were projected onto the environment which made it an augmented reality. And in this small video, we’ll see it in action. The robot, at the moment, is stuck in a corner, so the A for anger is quite high. But it moves around and eventually it will come out.
Still stuck. Now, assume, see? The curiosity getting bigger as it’s getting outside. When it really comes out, we’ll see the C for curiosity much higher and higher. Here he comes. And we’re very curious.
Third example is a humanoid robot. Now, in this course, we’ve mostly had wheeled ones, but legs can be more versatile, though it’s more tricky. And there’s the problem that if the robot falls over, how does it stand up? And here we’ve got Erez Ashkenzi’s project which basically, he’s got his humanoid-type robot, and he applies lots of advanced control techniques, much more advanced than ones in this course, to get the robot to stand. So here it goes. Starts to move around. Starts to try to get a foot on the floor. Yep. Yes, using its hand. It’s now got two feet. Uses its hand to push up. And there. Well done. So they were just three examples of final year project work.
And students get a lot out of their projects. Some, as in the first example, were very much related to robotic research at the time. And good projects even get written up in academic journals or appear at conferences.