How ASIMO Worked

and I'm happy to do it. Fun fact before we get into the Osomo story and what made it special and how it worked, the very first article I ever wrote for how stuff works dot Com once I was hired was how Osomo Works Now. I technically rewrote a couple of articles before, at where they wanted to test my abilities by giving me an existing article and saying could you rewrite this? Could you update it? But the

first full assignment beginning to end was how Osomo works. Um, if you want to read one of my even earlier ones, I did a rewrite on electronic voting machines, but Asomo was my first full article. The article is still up at how stuff works dot com. But it also was updated at some point by lee An Aubringer. So I didn't I wasn't consulted about that, because that's not the way it works. Typically when we write articles for how

stuff works. As you know, technology evolves very quickly over time, and so articles that you write will become outdated rapidly, and so we would frequently bring in other people to help update articles while our staff writers like myself would be on other new, big assignments. So, but this is all about AWESOMO. So what was awesomo. The name is

actually an acronym. It stands for Advanced Step in Innovative Mobility, and it's a robot that's intended to interact in a human living environment, meaning that Honda was trying to build a robot that could integrate into our lives, that could move around with human beings in a seamless way, and it was meant to be a humanoid robot that could interact with our environment similar to the way we humans do,

so it needed to have human like appendages. And it turns out this is a lot easier said than done, because a lot of consideration has to go into the design of such a robot. On casual glance. Awesome, Oh, looked like a child sized astronaut, complete with like a space pack and a space helmet. It's humanoid, has arms as legs as hands with fingers and thumbs, and like I said, the head looks like a like a space helmet that you would see an astronaut on the Moon wearing.

And it had it's batterying a big backpack that sat on its back obvious lee which gave it this kind of little estronaut looks very cute, and in fact, that was done on purpose. They wanted the robot to look friendly and approachable because it was meant to interact with people. UH. Over time, the design of Osumo evolved a little bit, and I'll cover some of that in this episode, but I just want to illustrate it here with the earliest version of the official Osomo, with the version that was

around just before they stopped production on it. In the first robot that Honda designated as Osomo, because there were some predecessors, UH, was born quote unquote on November two thousand. It stood one twenty centimeters tall that's just a little bit under four feet, and it weighed about fifty or a hundred fifteen pounds. But when it was retired in two thousand eighteen and Osumo's robots specs were a little different.

It stood one hundred thirty centimeters tall that's about four ft three inches, so it grew a little bit, and it weighed forty eight kilograms or about a hundred six pounds, so it lost a little weight. At top speed, it could dash at a respectable nine kilometers per hour, which is about five point six miles per hour. Now, it's not like it could go toe to toe with the T one thousand from Terminator two, but it was still pretty impressive. It was a robot that could run on

two legs, the first to do so. In fact, Honda chose the robot's height carefully. They wanted to design a robot that could work in human environments, so it couldn't be too small, but it also shouldn't be intimidating. They didn't want to make, you know, like a kill bot

two thousand that would be terrifying. The robot size was determined to be people friendly because it'd be large enough to operate environmental elements like door knobs or light switches, and about four ft three inches meant that it would also be around eye level with a seated human adults. So if you were seated down at a desk and Osimo walked up to you to let you know that there was a visitor at the lobby, then you'd be

looking pretty much I too camera with Osomo. It could also stand behind a desk and it would be at about the same height as someone who was seated behind a desk, So that made Osumo a potential receptionist, which in fact, Honda made use of Osimo as a receptionist at their headquarters. The robot also had several degrees of freedom. Now, the phrase degrees of freedom actually has a couple of different meanings depending upon what industry or context you're looking at,

uh including statistics. There's a specific meaning for degrees of freedom and statistics, But in mechanical systems it refers to the number of independent movements a rigid body is capable of. So an unrestrained rigid body in space just imagine you've got a cube of something and it's magically floating in the air. It has six degrees of freedom because it can move up or down like it can levitate straight

up or sink straight down. It could strafe left or right uh, or it could move back and forth toward you or away from you. Plus it can rotate around those axes, which would translate to pitch roll and yaw.

So that's six different degrees of freedom. The final model of Osimo had three degrees of freedom for the head, seven degrees of freedom for each arm, so I had two arms, thirteen degrees of freedom for each hand, two hands, two degrees of freedom at the hip, and six degrees of freedom for each of the legs two legs, which meant that Osimo ultimately had fifty seven degrees of freedom for the full robot, the one that was finally retired in that was an improvement of twenty three degrees of

freedom from its predecessor. Model had twenty three degrees more freedom than the previous version of Osomo. I'll go into further detail about the different sensors and systems Osumo had, but first I'd like to talk about the history of the robot itself, and to do that we need to travel back to night teen eighties six and what a

year that was, guys. I mean, you could go to the theater and you could go see what came out in eight six aliens you could see Ferris Bueller's Day Off, or you could see the greatest movie made of all time, Big Trouble in Little China. Also, Howard the Duck came out in as I recall music in six include stuff like Peter Gabriel's Sledgehammer or Peter STA's Glory of Love. Quite an amazing diddy there or cameos Immortal, classic word Up. But more importantly for our story engineers at Honda, we're

tackling a very challenging problem. How do you make a humanoid robot walk like a person does? So they started off slowly. First, they began with designing a basic set of robotic legs that could take steps. So their first robot model in this phase, which was only meant for research and development, this was never going to be something that they were going to market, was known as the e O and it was essentially nothing more than a pair of legs connected to a narrow set of robot hips.

So just legs and hips and that's it. And it was wired up and you had a control system that would tell it to walk forward. The robot was connected directly to those computer systems no wireless systems at this point, and the legs had to be very careful, very deliberate in their movements in order to maintain the robot's center of gravity on the soles of its feet. So every time it takes a step, it's it's removing weight from one foot. All of its weight is on its other foot.

It had to be very careful to move its center of gravity over the the foot that was still on the ground. That also meant that the robot had to continuously make adjustments to its balance as it moved the free leg forward to take a step. That made taking a single step a slow process, sometimes taking as long as twenty seconds for one step. Now, clearly that's way too slow for a robot to move around in a human environment, but it was important for the research and

development phase. The robot and all future robots i'll cover used servo motors for locomotion. Now, those are actuators. They're either linear or their rotary, which tells you the type of motion they control, and they allow for precise control of motion, including velocity and acceleration. A servo motor uses position feedback, which means the servo motor has to quote unquote no, it's position so that when it receives an incoming command to change positions, it can do so accurately.

So if I translate this into human terms, it's like saying, you put your right leg in, you take your right leg out, you put your right leg in, and then you shake it all about. Humans typically possess a sense that we call a proprioception. That's describes how our brains sense our bodies, how our brains know where our limbs are. This is why if you close your eyes, you can touch your finger to your nose, assuming you're not under

of the influence of something. Because we have appropriate exception, we know the location of each of our limbs, and therefore we can move those limbs from where they were to where they need to be. And that's what it's all about. But machines do not innately possess this ability. Engineers and computer scientists had to come up with ways to mimic it, and servo motors are part of systems that keep track of positions so that the overall system

can behave as directed. So if you tell your robot walk forward three steps, the robot can take that command and translate it into a series of smaller commands. If one leg is already placed forward, then the back leg is the one that's going to have to come up and take a step for locomotion to begin. For example, pretty much every degree of freedom has an associated servo motor. So as ASUMO or osimo, I should say, increased in complexity, it got more servo motors. Back to EO, The type

of walking EO could do was called static walking. Before it could take a second step, the robot had to be certain it's center of weight had shifted over the soul of the foot it had just placed down on the ground. This was a good start to working out the actual mechanics of limb motion, but it was a

far cry from the way organic critters move around. The team was gonna have to do a lot more research, and I'll talk about that in just a second, but first let's take a quick break to thank our sponsor. Engineers got to work analyzing the way humans, animals, and even insects walk. They studied hours of videos to understand what is going on from a physical or mechanical side

of things. You know, physiologically, humans don't maintain their center of gravity directly over the souls of their feet as they walk. The center of gravity as we walk can move around quite a bit, so designing a robot that could do this too was gonna be pretty tricky. The robot would have to take into account a lot of different factors that we kind of grasp innately once we

learn how to walk. That would include stuff like the robots fed its momentum, the ground it was walking on, whether it was level, whether it was flat, that kind of thing. One thing the team did was study human skeletons. They noted the location of the joints in the human legs and determined that the toes play an important part of walking as they helped guide us in the way we support our weight from step to step, so their robots would need to be able to do a similar thing.

The robots were also going to need degrees of freedom similar to what you could find in a human ankle, knee, and hip joint, so the engineers also studied walking humans to determine stuff like the range of motion every joint should be able to replicate, where the center of gravity should be for every leg, how much torque should be exerted on leg joints, and also the sensors that would be needed to replicate how we humans sense, stuff like the speed of motion and the impact of our foot

hitting the ground. All that would be very important so that the robot would be able to walk in a stable way and not just hop around or fall off of its feet or otherwise have some disaster occur in the engineers designed E one. This was the first of three robots, the others being E two and E three that the engineers designed in an effort to move from the static walking model to what they called fast walking

Like EO. These robots were essentially legs attached to robotic hips, and maybe you could argue as also sort of a rudimentary torso, but there were no arms, there's no head, so I was about it. Uh. Each design looked a little bit more sophisticated than its predecessor did, but they were all very almost industrial looking kind of robots. They would would not take steps as painstakingly slowly as EO did. They moved a bit more naturally, which for humans involves

shifting our weight forward and leaning into a step. It's almost like we're about to fall, right like when we take a step. It's almost as if we're leaning forward and we're gonna fall if we don't catch ourselves, and then we move a foot forward and we do catch ourselves with our foot, and then we keep leaning forward

and we catch ourselves again with our foot. So you can almost think of walking as consistently nearly falling over in a way if you're looking at it from a robotics way where you're trying to figure out how to design a robot to do a similar thing, So it involves a lot of almost falling and catching yourself. While the robots in this phase were considered fast walkers, fast

is a relative term. The E two, for example, the middle of the three models, was clocked at a top speed of one point two kilometers per hour on flat surfaces, which is about three quarters of a mile per hour. By comparison, your average humans walking speed is right around five kilometers per hour or three point one miles per hour. So these robots weren't exactly burning up the track walking around the work on E one through E three stretch from seven to and then the team moved into the

next as new wave dance craze. Anyways, it's still rock

and roll to me. I don't know what happened there, it's all on my notes to What I meant to say is that they began developing the next generation of walking robots that would be E four, E five, and E six, and this wasn't an effort to create stabilized walking, which meant the engineers wanted to create robots that could remain stable while walking on a variety of surfaces, including stuff like slopes or stairs, that they would be able to adjust their steps and be able to keep their

weight centered so that they didn't fall over. These robots, like the one through E three also looked like an armless torso attached to legs, so still didn't look very much like Awesomo. At this point, the E four through E six models began to incorporate three areas of control to achieve stabilized walking, and they are floor reaction control that refers to the ability of the robot to absorb floor unevenness through the soles of its feet in an

effort to maintain a firm stance. So the engineers had to build sensors into the feet of the robots so that the robot could gather information and about the floor and then process that information in a way that was meaningful and then adjust its stance to give the robot the best chance of not pitching over. Next would be the target zero moment point control, which is a fancy way of saying the robot needs to be able to balance itself, so the zero moment point refers to balancing

different forces in order to maintain posture. Those forces include stuff like gravity and walking speed that falls into a category called total inertial force. The other force that occurs is when the robot's foot connects with the ground. That's called the ground reaction force. And you want those two forces to cancel one another out in order to maintain posture. Also, the robot has to be able to detect when it

is unable to stand firmly. So first you have to incorporate sensors that can detect and imbalance in the robot. Then you have to figure out what to do without you know, how do you address and imbalance. So with these robots, the engineers designed a system which will allow the robots to make adjustments to its upper body and they would shift their upper body around to act as

a counter balance. So if it's since it was going to fall forward, it might shift its upper body backward to counter that action and hopefully remain upright as a result. The third area of control was called foot planting location control. This system engaged once the ZMP control had activated, and this system would determine the length of the robot step to catch the robot and make certain it remains upright.

So it's all about maintaining the proper relationship between the position and speed of the robot body with the length of the steps it takes. Now, up to this point, all the robots have been prototypes to help engineers understand the fundamentals necessary that would be needed for basic walking. The next stage involved building robots that had arms, hands, and a head, and that wasn't just for aesthetics, although that did play a part in it, but it's also

for locomotion. We use our arms and our body in our head while we're walking. If the engineers wanted their robot to move like a human, they were going to need to incorporate those elements as well. Plus, if they wanted it to interact in human environments, they wanted it to look not terrifying, So giving it arms in the head was probably a step in the right direction. From the team built another series of three robots. These were designated P one, P two, and P three, and all

of them were humanoid. They were all taller and heavier than Osumo would be, and this was when the team was still working out the physics and mechanics of humanoid walking, so they were more concerned with getting those elements right rather than producing a robot that would be suitable for human use, so again these were never intended to go into the workplace. The P one humanoid robot was one

five centimeters tall that's about six ft three inches. It's a big robot and also weighed into a hundred which is about three eighty six pounds, so stats like that. It could have wrestled for the w w E. Now, clearly that type of robot would be too big and heavy and potentially dangerous for a human environment. If it lost its balance and fell, it could cause a serious injury.

But it was one of the first of Honda's robots in this line to have arms and sort of claw like hands, and engineers worked on coordinating arm and leg movements and programmed the robots so it could operate simple things like light switches and door knobs. And pick up various objects, and coordinating all of that was also another big challenge, although to be fair, the walking and running was probably the biggest of the challenges they faced at

that point. The P two robot was the first self regulating, two legged humanoid robot walking robot, i should say, and it first started strutting its stuff in December n This robot was the first to have a computer system incorporated directly into the robots design. Previous robots had been controlled by computers through a wired connection. This one was completely wireless. The P two had a battery, It also had a wireless radio, had motor drives, it had its control computer,

and more systems on board. Operators would say commands to the robot through a computer, it would beam the commands over. The robot would receive these commands wirelessly, and then they would process the commands and the robot would then do whatever it was supposed to do, including pushing carts or climbing stairs. This robot was a little bit shorter than the P one. It measured a hundred a D two centimeters, so it was just a hair under six ft tall.

But putting all that on board processing capability onto the robots uh skeleton meant that they added a lot of weight to it, so it was a hefty two kilograms or nearly four hundred sixty three pounds. I stumbled there for a second because in my notes, just a glance point the curtain, I wrote two ten kb. That's two bites, No two kilograms, silly typo. The P three, which was created in September, was much shorter and lighter than its predecessors.

It stood one sixty centimeters tall that's about five ft three inches, and it weighed in a relatively sevelled hundred third kilograms or two seven pounds. The team was able to decentralize the control system for the P three, which helped remove a lot of that weight that it was carried around by the P two. Now, those robots helped the engineers put together the information they needed to create

the first robot to be called Asomo. This robot would be smaller, it would be lighter, and it would feature a design that was meant to make it look friendly and playful. So next I'll talk about some of the tech that was used to make Osumo work and how it became the first humanoid two legged robot. To run. But first, let's take another quick break and thank our sponsor. Asomo represented a big breakthrough in creating a robot that

can walk like a human can. For one thing, the engineers developed what they called intelligent real time flexible walking or I walk for Osomo. So Osmo can shift its center of gravity while going through a turn, and that allows it to make a turn in a gradual curve. It's like it's it's leaning into the curve, which is a big deal because earlier robots the only way they could turn is they would actually have to stop moving

and then they would sort of shuffle in place. They would stand on one foot, lift their other foot, turn it slightly, put their other foot down, lift up their first foot, and put it in parallel with the second foot, and then they have to keep doing that over and over and over again until they gradually we're facing in the correct direction. So Osomo was able to do this in a much more fluid way, one that was much more human, which is very important if you're gonna have

it moving through human environments. It can actually calculate the amount of momentum it will need to get through a turn and shift its weight to help compensate for all of that. Osomo is also the first humanoid two legged robot to run. I mentioned that earlier, and by run, I mean Osomo can move forward at a slightly accelerated pace and in such a way that at some point during its gait both of its feet are off the ground at the same time. That's how they define running.

It's not by super top speed, but rather that at some point in the in its stride, both feet are actually off the ground. It happens for less than a second, but it means that Osumo moves around with a kind of like a little hopping motion. Some people have said that it looks like whoever is inside the space suit needs to go to the bathroom, but it is running.

And it's actually a big deal for robots because when the robot has both of its feet off the ground, it no longer has any information about its balance with regard to the ground. Right it's actually completely clear of the ground for even a split second. Osumo has to be able to maintain it's balance and it's weight so that doesn't spin when it goes off the ground. It has to be able to plan a foot down for the next step and be uh, firm enough so it

can continue it's run. And it has to do all of this without having any touch to the ground at all. And it's a pretty complicated problem to solve from an engineering perspective. How do you get a robot so that when it leaves the ground, it maintain aims it's orientation so that it doesn't just twist out of the way and then come crashing down. One thing I think is interesting about Osmo is that it combines pregenerated environmental models

and the ability to recognize things within an environment. Now by that, I mean you can't just PLoP Osomo down into a brand new space and expect the robot to seamlessly navigate through various interactions. Osumo is not autonomous. It's

not an autonomous robot. It cannot operate all on its own. Instead, it relies on a combination of programming, an operator who can run Osomo from a computer sort of like a remote controlled vehicle or both, in addition to its own ability, so it can respond to things like verbal commands and gesture commands. So it could do all those, but it can't do stuff on its own. This does not mean

it's not an impressive technology. This is very impressive. If you were to take control of Osumo and make it walk forward, it could actually adjust its own steps to meet your commands, which might include move having a leg a certain way to avoid an obstacle while still navigating to the location you guide it to. So, in other words, Osmo can make small decisions like where it needs to move body parts in order to continue to fulfill whatever the command was that was given to it. But it

can't decide to do something all on its own. It's making these smaller decisions about minute stuff within the context of a larger command. To do things like navigate stairways. You typically would program Osumo to have a working knowledge of the environment before having it moved through the room. That helps Osumo's navigational systems as well as it means the robot knows where things are supposed to be and can compare where things really are against that model in

order to make decisions. So, for example, let's say you have Osumo walking around a hotel lobby greeting guests, and one morning, some guests have moved a few of the chairs so that they can sit together. But the chairs are no longer where they used to be in Osomo's little preplanned man model of the hotel lobby, and if it only worked off of that pre program model, then it would try and walk around chairs that weren't actually there, or it would bump into chairs that had been moved.

But Osomo can also use its sensors, which I'm going to talk about in the second, to locate obstacles and navigate around them within the context of the room, and knows how far out the way it can go in any given direction. To plot out its path, typically you would also mark a room with certain types of markers that Osomo can detect, and that gives Osumo more precise information about where it is in context of other things

inside the room. Generally speaking, Osumo tries to take the pathway that will require the fewest number of steps or the least amount of work. And it's not that Osmo is lazy, but rather it has a fifty one point eight volt lithium ion battery and that's good for about one hour of operation, so it takes about three hours to recharge it. So Osumo wants to make the most

out of its brief moments of wakefulness. So note to self, if I ever get an Osumo, make sure I also pick up a couple of extra batteries, so I always have one charged. Those batteries, by the way, account for nearly six kilograms of Ozumo's weight, or about thirteen pounds. So let's talk about some of the sensors Ozomo has

to help it navigate an environment. In order to maintain its balance, Osmo has a gyroscope, and gyroscopes are devices that measure or maintain rotational motion, and they have some interesting properties that, on a casual glance, seemed to defy common sense. A gyroscope works on the principle of angular velocity around an axis of rotation. So imagine you've got a bike wheel and you have an axle, and the

bike wheel can rotate freely around the axle. So you're holding it up right with one hand on either side of this axle. You're holding it up in front of you, and your friend gives the wheel a really good spin. You're holding it vertically. You then attach a string to one side of the axle of this upright wheel, and your hand is still on the other side. Now, what would happen if you were to let go of the axle? And you might think that, well, the wheel is just

gonna flop over horizontally, but it doesn't. The wheel, as long as it's rotating rounds axle will remain upright, and that effect is called precession. A spinning gyroscope will uh is stable, and once the gyroscope is spinning, it has a tendency to remain in the same orientation, and any force applied to change the orientation of the spin axis

is met with a resistive force. So let's say you're still holding both sides of the axle of this bicycle wheel and your friend gives it a really good spin. You're holding it vertically, and then you try to turn the wheel horizontally while it's still spinning. You would actually feel resistance to this. This is what actually makes it easy to ride a bicycle. Once you really get going, the rotation of the wheels along their spin access will help keep you upright. In addition to the gyroscope, Osomo

has an accelerometer which measures acceleration. Acceleration refers to a change in velocity, so that could be either in speed or direction or both. Osamo also has a six as to six access force sensors that's for detecting the direction and amount of force that the hand ends encounter. Uh. They also actually have two more for the feet. I forgot about that, So technically it's got four six access

force sensors. Osmo also has cameras to provide a stereoscopic view of its surroundings that allows Ozomo to judge the depth of a scene and determine which objects are close to it, which one versus which ones are further away. The systems on board Osmo also have facial recognition capabilities and allow the robot to recognize objects that are in motion, and it can also respond to gesture commands like waving. So if you wave at Osmo, it can wave back

at you and recognizes that as a gesture command. Osmo also has environment identifying sensors, including ultrasonic sensors that can detect obstacles that are up to three meters ahead of it, including glass. Because you know, it works off echolocation, it's not optical. They are also laser sensors and infrared sensors to help the robot detect the ground. Osimo frequently navigates by referencing those markings on the ground I mentioned before.

The infrared sensor can detect those and that tells Ozomo it's in the place, or that it needs to move in a certain area that may not be in the location it thought it was. The ground sensor is located at the base of Osomo's torso, and there are ultrasonic sensors in an array both on the front and the backside of Osumo. Osmo also has three microphones that allow it to detect noise and determine the origin of the noise,

as well as to receive voice commands. And for a while, you could meet Osumo at Disneyland Interventions in an attraction titled Say Hello to Honda's Asumo, and I did. I got to meet Osmo. I just went to watch the live show and that's all I was gonna do, is was was just watch it and walk away. But a cast member was talking to me and I mentioned that I had written an article about how Ozomo works, and they said,

would you like to meet Osumo? And I said yes, I would please, and so they introduced me and I got to meet Osumo. Osmo, by the way, you can still be seen at Disneyland's Autopia, but Osumo itself is

no longer in production. Instead, Honda plans to incorporate the technology and discoveries made during developing the robot into stuff like the unicub device or the walking assist Harness and Honda introduced four new helper robot concepts at CE, and all of them seem to incorporate some elements of Osmo's design in them. So Osumo lives on in a way, but in new products, though the form factor of the childlike space man appears to be a thing in the past.

So farewell, Osumo. You're really good to me. I appreciate you giving me my first writing assignment here at how Stuff Works, and thank you Cat for suggesting the topic. It was a lot of fun to go back and revisit this and watch videos of Osimo running around and sometimes toppling over. It's sad to see, but you know, technology doesn't always work the way you wanted to, and it's good to remind ourselves of that from time to time.

If you have any suggestions for topics I should tackle in future episodes of tech Stuff, write me and let me know. The email addresses tech Stuff at how stuff Works dot com, or you can drop me a line on Facebook or Twitter. The handle of both of those is tech Stuff H. S W. Don't forget to follow the show over on Instagram and I'll talk to you again really soon. For more on this and thousands of other topics. Is that how stuff Works dot com