The Spot robot developed by Boston Dynamics possesses incredible stability, maneuverability and versatility. For example please have a look how this robot recovers in case of a fall. In this article I will try to unveil some of the secrets related to Spot robot’s design, walking pattern, fall recovery and robot vision.
Why backward bend legs?
The leg angle of the spot robot is interesting. It’s a backward bend design. Let’s explore the backward leg angle of the spot with another angle which is forward bend design. Which leg bend design can easily climb the stairs? The forward bend design or the backward bend design? The forward bend robot can collide with the stairs due to its bend position(Fig 2a), but the backward bend leg gives it a clear advantage. Backward bend design climbs the stairs easily as shown below(Fig 2b).
A funny reason can be, why Spot’s engineer chose a backward bend design is because backward bend robot can give you a handshake quite easily. I will explore the real reason towards the end of this article. Let’s achieve this amazing spot robot design starting from the simplest possible design.
The most simple design
In the simple spot design(Fig 3a), each leg needs two motors for its operation. This clearly shows the details of the motor connection at the hip joint. A planetary gear set is used in between the motor and upper arm(Fig 3b). This is for torque multiplication.
Can you tell me how a spot robot operates the upper limb when you power the motor? Now, let’s me check how the knee motor is connected. The stator of the knee motor is fitted with the lower end of the upper limb. The lower limb is now connected with the rotor of the knee motor. It’s an interesting arrangement. The upper limb is supported by a motor and this limb is supporting another motor(Fig 3c).
Using these 8 motors as shown in Fig 3a, you can achieve any angle for the 8 limbs. We can now attach a controller to the robot and vary these angles randomly. Clearly, the robot won’t be able to make any meaningful movements if we simply feed it random angles. The Spot robot may even get stuck. We can imitate nature and resolve this issue. Here, we are using the recorded angles of the limbs of a dog at different instances while it walks. Now, I will feed this agles data to the robot, with these angles the robot moves forwards smoothly (Fig 4).
However, with this design the Spot’s battery charge quickly depletes. Let’s analyse why this is happening. When the hip motor rotates, it has to lift the weight of the upper limb, lower limb and knee motor (Fig 5). During the leg lifting operation the various forces are acting on the motor.
A trick with the knee motor : saves energy consumption
How we can reduce power required for this lifting? We just need to moving the knee motor closer to the hip motor(Fig 6). Since the knee motor weight is closer to the hip motor, the torque required drastically reduces. However, how to control the lower arm? It is far away from the knee motor. The answer is obviously a mechanism between the knee motor and lower limb.
Before understanding this mechanism, let’s make a small design change so that the torque requirement will slightly be reduced. Replace all bulky legs with thin – carbon fibre legs. Now, let’s get into the mechanism of the Boston dynamics engineers came up with to operate the knee joint.
Ball screw mechanism
To achieve ball screw mechanism we must extend the lower limb behind the knee joint. Now, if the motor can give a linear motion to the tip of the lower limb, it will obviously rotate. The best way to produce linear motion from a motor is a ball-screw mechanism(Fig 7). For clear understanding let’s consider bolt which is directly connected with the motor. When the motor rotates the nut and bolt rotates together as a single piece. There is no linear motion. However, when you arrest the nut you get a linear output motion from the nut. In the robot ball-screw arrangement also there won’t be any linear motion if the ball nut is not arrested. This is why a carrier is used. The carrier arrests the rotation of the ball nut and the ball nut moves linearly. Now the lower limb is controlled quite accurately using a motor far away.
All there is left to do is to add some absolute encoder and torque sensors, and this spot robot is ready to strut flawlessly and confidently into its next challenge.
How does the spot robot recover?
At the beginning of this article, we discussed how the robot recover from fall. The solution to recover from this position is an interesting design change. Bring one more motor in and connect its rotor with the body of the hip motor. Looks like a crazy arrangement, right? This is what happens to the hip motor when the tilt motor operates(Fig 8). When you operate all four new motors together, the legs widen, and the robot appears to be much more graceful.
Let’s see how it recovers from falling down position. The importance of the third motor we introduced – the tilt motor is pretty clear from this visual. Our robot is back on its four feet(Fig 9).
The walking pattern
The walking pattern of the spot is known as trot motion(Fig 10). Our beloved furry friends, dogs and cats, walk this way. Our Spot robot can do one more interesting walking pattern: crawling. Obviously, the robot will select this kind of pattern when stability is of utmost importance. The Spot can also be fitted with an arm on its mounting rails. This arm includes six motors for 6 dof and a gripper to grasp objects. To detect objects, a camera is placed inside the gripper. With this arm, it can easily pull the levers in industries or open doors in a people-filled environment.
Why does the Spot Robot has visual camera on its back?
The Spot robot we have developed is really smart. However, it is interesting to know that such a smart robot will fail during a simple stair climbdown case due to the similar reason we explained in the introduction of this article. We have already seen that the Spot robot’s backward bend legs help it to climb up without any collision. This means the backward bend leg will collide with the stairs while climbing down. Here’s a trick to solve this issue. Just program the robot to walk backwards during the climb down operation. This is why the Spot robot has a set of visual cameras on its back(Fig 11).
Why the backward bend design?
Now the genius part of the this article – the main reason behind the selection of the backward bend design. Mr. Jackowski, the chief engineer of the Spot robotics project, went ahead with a backward bend design for the Spot robot. We are able to walk because our legs derive a reaction force from the ground(Fig 12a). The same is the case with the robot too. In a good robot design both the motors should resist this reaction force or participate actively in robot motion. Which design has this quality? To find out the answer, let’s do an experiment, but instead of the electric motors we are using torsion springs(Fig 12b). What do you think, in which design both these springs will get compressed?
I hope most of you have given the answer B. In the design A when I apply the force the spring 2 is not at all getting either compressed or stretched. But, in the design B, both these springs are going for a compression(Fig 13).
This shows that in the backward bend design both the motors will play an equal role. This is exactly why the Boston Dynamics engineers selected backward bend design for the Spot robot. You can easily find the reason behind it with this torque analysis. Would you like to be a robotics engineer? To start your career in robotics please check out our detailed course on robotics. Thank you.
3 Comments
Nicely explained about this application of robotics.
Hope to see more articles in future.
All the best.
I am happy to say that your masters are completed from IIT Delhi, it’s a very professional place.
Sir, I have one request can you please guide us to make this robot with other concepts.
We are working for agricultural purposes.
Which will help to farmers a lot.
If you are interested please let me know.
Good article, appreciate the details