Skip to 0 minutes and 17 seconds Baxter, in contrast to the little mobile robots that we’ve seen so far, is the latest version of a more manipulator type robot. Whereas a traditional manipulator is a great big powerful device, not safe to be anywhere near, Baxter has been specifically designed that humans can work in the same environment. And in fact, you can grab hold of Baxter’s arms and move it around. Which, for instance, you can use to programme the Baxter to perform a particular task.
Skip to 0 minutes and 54 seconds And Baxter has been designed to interact with humans, as we shall show in this little demonstration. If I get closer, as he’s asking me, like so, get close enough, puts out his hand. And I’m Richard. Nice to meet you. So if we look at Baxter as a whole, we can see that it has its two arms, each of which have seven degrees of freedom, which means it’s almost as flexible as a human arm and therefore, as long as it’s within reach, the robot can pick up any object.
Skip to 1 minute and 33 seconds If you think about how you as a human would try to pick an object up, you’ve got to move your arm until it’s in the right place, which involves having your different joints to be in the correct position. Well, for Baxter, it has a series of motors which are actuators. You’ve got the gripper at the end, which you can move in and out. It’s got a wrist joint. It’s got an elbow joint. It’s got a shoulder joint, just as we have. And so there are motors which are driving each of those joints. There is control associated with them for controlling the position or the angle of each joint. You can also specify how fast. So you have some velocity control.
Skip to 2 minutes and 17 seconds And also what sort of force is produced at each of the joints. That is taught control. And it’s quite important because Baxter’s designed to work for humans, that it isn’t what is prevented from turning around to rapidly, providing too much of a force, because it could knock you over. So taught control is important for Baxter. If Baxter is to be able to pick something up, it needs to know where its gripper is. For each of its two arms, there are a variety of different senses. In order to detect, it’s got a camera at the end of the arm. It has infrared sensors, which like the mobile robots are used to detect distance.
Skip to 2 minutes and 57 seconds In this case, how far is the gripper above the table? It also has accelerometers allowing it to measure force. It also has a set of ultrasonic sensors around the head and there are flashing lights to show you that they are there, which can tell whether is a human quite close. Which again, is part of this built-in safety mechanism.
Interacting with Baxter
The original industrial robots were large devices capable of moving very rapidly where it was important for safety reasons that humans kept out of the way. The robots may have been put in a cage as a result.
The Baxter robot is different – it has been designed specifically so that humans and it can be in the same place and can interact with them safely.
This means that Baxter has to be able to detect where humans are and so attempt not to bump into them. It uses its patented ‘series elastic actuators’ and ‘force detecting sensors’ which mean its robot arms are ‘compliant’ so that when they do hit a human or other object they do so safely.
We have seen the robot picking up balls in the previous Step, and that requires a relatively complicated program. If it is to do some other task, a different program is needed. It takes time to write such programs and get them right. How do you decide what Baxter should do?
To address this issue Baxter has a nice feature. It stems from the fact that a good way of teaching someone something, is to show what you want. So, if Baxter is to pick up balls, one way of doing that is to grab hold of one of its arms and move it until the gripper is in the right place – showing Baxter what it should do.
In this video you see Richard teaching Baxter to move with one arm, and the other arms does the same actions – a great example of human-robot interaction.
© University of Reading