Making it Real 

It has long been a dream of roboticists to achieve something that looks and feels like real human walking from their elegant machines. It’s clear that a lot of progress has been made but even the most casual observer can see that robot walking is still different than human walking.  Though, interestingly, humans seem to enjoy imitating robotic movements.  (Ref: 

Smooth Movement 

To understand what happens during walking (aka bipedal movement) it is instructional to watch robotic movements with the understanding of the complexities of how humans are able to do it so smoothly.  Let’s start with the simplest structure that embodies the challenge of walking: the inverted pendulum.  A regular pendulum is just simply a weight on the end of a string.  It is inherently stable with the center of gravity (CG) of the weight (the “bob”) well below the attachment point of the string.  It will always point straight toward the center of the earth. 

Maintaining Balance 

An inverted pendulum has the bob at the top, and because a string will collapse in this orientation it needs to be replaced with a stiff, lightweight rod.  This arrangement is inherently unstable.  It is virtually impossible to balance this type of pendulum with the bob on top of the rod. This is the human equivalent of balancing your center of gravity (which is about 2cm below your navel) on one leg, while your other leg swings forward to take a step.  Once you have leaned forward and your CG moves past your foot, you are, in essence, falling – just like our inverted pendulum.  To get an idea of just how complicated and unstable this movement is, stand up and balance on one foot for about 30 seconds.  Focus on your ankle and you will readily notice how much it is moving around, constantly adjusting and trying to keep you from falling over. 

Do the Robot Shuffle 

Now for comparison take a look at this video of a robot doing some human-like walking.  (Ref:  Watch carefully and you will see that it gains some stability from its squarish feet which eliminates some of the ankle wiggle that is inherent during steps (as you experienced in the experiment in the last paragraph).  Also note, that the small yellow and black target right at the robot’s “waist” is able to pivot, imitating the human pelvic rotation. This reduces movement of the CG from side-to-side, and hence reduces wobble. This helps make its movement appear more lifelike and fluid. 

How Feedback Works 

Text BoxText Box

It is critical, whether robot or human, to receive feedback about your movement to avoid falling. Both have a way to sense gravity which helps orient the body to vertical and by inference, horizontal.  These tilt sensors (called accelerometers) are MEMS (Micro-Electro-Mechanical Systems) devices that respond to earth’s gravity. An example is shown in Figure 1.  The yellow part is known as the proof mass which is attached to a rigid frame via some springs.  You can see that it is interleaved with a series of white “fingers” and the device is constantly monitoring the capacitance between the moveable yellow fingers and the fixed white fingers.  You could image if this sensor was sitting flat on a table that it would have an output value of zero, since there is no applied force to make the yellow proof mass move.  However, if you were to tilt it so that the left side was aimed more toward the ground, then the yellow part would slide to the left constrained only by the springs.  All of the gaps between the yellow and the white fingers would get smaller on the downward side and wider on the upward side.  This will cause an imbalance in the capacitance on one side versus the other.  To rebalance this, an electrical charge is injected to deflect the yellow proof mass assembly back to the neutral position.  The amount of charge needed to do that is proportional to the deflection of the spring-mounted mass. By measuring the charge, the applied acceleration is known for the robot. 

And for Humans . . . 

Humans solve this problem in an analogous way. We have structures in our inner ears that respond to horizontal and vertical movement: the Utricle and the Saccule.  Small calcium/protein stones are attached to cilia that sense shear forces during acceleration, enabling them to distinguish both magnitude and direction of changes in movement. 

But Wait, There’s More 

There are two more factors that come into play to create the smooth and versatile movement of humans. The first is proprioception: it is the body’s innate sense of movement, action and physical space.  It’s sometimes called muscle memory or kinesthesia. This sense of where you are in space is what allows for athletic endeavors like pole vault, downhill skiing, gymnastics and similar activities.  If you have to actually think about these activities, you wouldn’t be able to do them.  The body can “know where it is” while going through these motions in a way that is virtually instantaneous.  There is no robotic equivalent. 

The second thing is that humans use their sense of sight to help with balance while moving.  Robots can, too – but in a little different way. Their vision systems are mostly designed for collision avoidance.  The human sense of sight is combined with information from the inner ear balancing mechanism to keep the body upright.  For a great illustration of this.  Go back to the inverted pendulum demonstration from before and try standing on one leg, again.  This time, after a few seconds, when you feel pretty stable, close your eyes and see how much more difficult it is to do a single leg stand with your eyes closed.  Don’t be surprised if you fall over. 

We’ve just touched the surface of the balance problem in robotics, and it’s useful to understand the human analog version of balancing to grasp the complexity. The future is unknown, but if we look at the history of robotic development, odds are that robot mobility will be solved.   It probably won’t look like the human solution, but then who’s to say it won’t be better?

Leave a reply

Your email address will not be published. Required fields are marked *