ROBOTS: FastRunner: A Robotic Bird Designed to Run Instead of Fly?

9 January 2014

In a DARPA-funded project, MIT was contracted to design and build a robot that runs fast and can walk through rough terrain.  So, they’re designing a robot to imitate . . . a bird? 

Yes, a bird. 

The world’s fastest running animal is a bird.  A flightless bird.   The ostrich.  In fact, the ostrich can run so fast, it’s probably never felt the need to fly.

DARPA has funded the joint effort of MIT and the Florida Institute of Human and Machine Cognition (IHMC) in a project to develop a robot that walks and runs.  Past DARPA-funded projects have resulted in the quadrupedal robots, BigDog, Robo-Cheetah and the Wildcat.  But the end result of this latest effort will be the first robotic biped in the DARPA arsenal.

Robo-Ostrich or FastRunner

 

Robo-Ostrich is designed not just to walk, but to run and run fast.  Although the first full prototype has yet to be designed, the working computer simulation has legs and is hitting speeds of 27 mph.  Impressive considering this is about the speed of the fastest human runner – in a hundred yard dash.  But this robot could sustain that speed indefinitely.

The designers, however, aren’t satisfied with a mere 27 mph and are hoping to, eventually, develop a ‘bot that will reach a speed of 50 mph.  And the 50 mph mark would be another milestone on two counts.  First, real ostriches clock no higher than about 43 mph. And, second, real ostriches are the fastest land animals on earth.  So, the 50 mph Robo-Ostrich would not only beat the real bird, but would also beat every other land-based animal on the planet.

Although this ‘bot is formally named “FastRunner,” it has come to be known, informally, as Robo-Ostrich.  Why?  Because the only way to develop a robot that could run as fast as an ostrich was to build its legs to as closely imitate the legs of a real ostrich as possible.  And if your ostrich robot is really going to perform like an ostrich, it needs the legs and, more or less, the rest of the ostrich body to go with it. 

By the way, this is called biomimicry – designing a technology to imitate nature in order to solve a complex human problem.

Indeed, everyone is so excited about the Robo-Ostrich’s performance that it’s easy to forget that this robot doesn’t really exist.  Right now, the ‘bot is a computer simulation, which is only about 40% complete.

Robo-Ostrich / FastRunner Simulation

However, this isn’t the “damper” it once was because modern computer simulations are remarkably good.  In fact, modern computer simulations are so good that they quite precisely predict the performance of the real things they simulate.  So, if you can “get it right” on the computer, you can break out your hammer and wrench (figuratively speaking) and start building.  But the building phase for Robo-Ostrich is still “a ways off.”

The development of Robo-Ostrich is particularly significant because this robot’s working legs will incorporate advanced technologies to maintain the robot’s balance.  In the past, designers attempted to build complex systems into robotic legs that would monitor and respond to every variation in movement on every type of terrain.  This required large, on-board computers, complex programs, and equally complex mechanics to control every aspect of simulated walking and running.

However, a new non-linear approach is being used in the development of Robo-Ostrich.  Although complicated to develop, the new system will be of a much simpler design.

To oversimplify, imagine your automobile with computers in each wheel monitoring every bump and, then, commanding the suspension system to precisely respond in order to compensate for each disruption. 

Readers familiar with automotive suspension will furrow their brow and ask, “Why?”  For almost a century, automobiles have used a spring that flexes when the wheel rolls over that speed bump (for example) and, then, returns the chassis to its original position – no computers required.

Very, very roughly, a similar set of principles are being used to develop the mechanics of Robo-Ostrich’s legs.  Though much more complex than an automobile suspension system, the goal is a relatively simple, self regulating balance mechanism that allows the ‘bot to maintain its balance as it walks over uneven surfaces.

The real ostrich can grow to a height in excess of 9 feet and weigh as much as 250 pounds.  In other words, you wouldn’t want to meet the real bird – in a bad mood — in an ally — at night.

Robo-Ostrich, however, will only measure about half the height and weight of the real bird.  This relatively “petite” size and weight produce an intended advantage. The lighter weight makes the robot faster and lowers its power requirements extending its range.

The two legged design has distinct advantages over the past quadrupedal models.  Not only is a two-legged robot lighter, but its movements are more flexible allowing it to, among other things, “get through narrower spaces” and maneuver more easily around obstacles.  With such a flexible build, this robot, like other “be-footed” robots, is designed to negotiate rough terrain that would defy a wheeled-vehicle like a jeep.  Even on irregular surfaces, the finished ‘bot is expected to run (or walk) at a speed of 10 mph.

Although still in the design phase, everything about the development of Robo-Ostrich seems to be on track.  Still, it will be a while before we’ll see the actual two-legged bird ‘bot walking and running . . . but not flying.

Thursday 9 January 2014

GCLM5444HOxenia

ROBOTS: Fast On Their Feet -- Robo-Cheetah & The Wildcat

6 February 2014



            Creating a “legged” robot is one thing.  Making it run instead of walk is another.  But, then, there’s yet another challenge.  Can you make it run fast?

            Developed for DARPA by Boston Dynamics, the robo-cheetah’s claim to fame is its speed.  Modeled after the real-life cheetah, this robot boasts a “cat-like spine,” which “flexes and extends” with the robot’s galloping stride. And it gallops -- “constantly tipping forward, falling, and regaining equilibrium with every step.”  After the development of the first prototype, in 2011, it was showcased running at speeds of up to 18 mph by March of 2012.  By September, it clocked 28.3 mph – faster than the fastest human runner in a hundred-yard dash.

            Of course, with all the excitement, Robo-Cheetah still had a couple issues that needed to be ironed-out before it could go bounding across a battlefield.  It was running at high speeds, but it was only running on a treadmill.  Still, it was about ready to jump off the treadmill and, at least, onto flat ground. 

            Robo-Cheetah's biggest issue was that it was still “tethered” by a power cord.  In other words, it had to be plugged into a wall socket to get the juice it needed to move.   There’s no portable power pack for this ‘bot that can store enough power to let it run free.  Portable power supplies are a big issue in robotics and one of the biggest challenges to maximum performance. 

            There’s a tradeoff.  You need enough power to allow the ‘bot to operate for long stretches of time.  You, also, need a power pack light-weight enough for the ‘bot to carry.  But, when you make the power pack light enough, there’s not enough power to run the ‘bot.  And, when you make a power pack with enough power, the pack (and ‘bot) become so heavy that, now, . . . the 'bot's short on power -- again.  It's like running in a circle. 

Robo-Cheetah 

            But, soon, there were more than technical challenges – there were challengers.   The first competitor was MIT. The Biomimetic Robotics Lab at MIT, also under the sponsorship of DARPA, was, and is, working on its own version of the robo-cheetah.  MIT is trying to recreate the running movement of the real cheetah.  They’re more public with their work.  The MIT website shows their version of Robo-Cheetah.  Their robot can’t run as fast as the Boston Dynamics model, but MIT’s model boasts a “highly efficient leg motor, imitation tendons, and a responsive tail.”  With these improvements MIT’s Robo-Cheetah has a rhythm and movement completely different from other four-legged ‘bots. 

MIT's Version of Robo-Cheetah 

            MIT's Robo-Cheetah, also, “will” run on a battery (but it, too, is still plugged into a wall-socket).  Unlike the other Robo-Cheetah, MIT’s uses a surprisingly simple and more effective way of regulating its leg motion – one without the usual sensors and complicated computer feedback-loops that were, and are, still a common part of robotic technology.

            But what’s so important about imitating a real cheetah?  The robo-cheetah is one of a group of DARPA-funded projects applying the concepts of biorobotics.  To meet DARPA’s requirements, drones must be built to perform more like . . . wildlife.  The term “biomimetics” or “biomimicry” is used to describe the development of technology designed to imitate and replicate the activities of biological systems and organisms.   But, why imitate nature?  Well, “if you want drones that work in a particular way, and the only known example of such performance is a biological organism, you’ll either have to imitate the organism or forget the project altogether.”

            The need for walking (rather than rolling) robots is a prime example.  The jeep took “a basic automobile and raised the center of gravity, increased the size and scale of the automotive suspension system and produced spectacular off-road performance for a machine with wheels.”  But the wheel, itself, was limited.  Human beings, horses, mules, and dogs can all travel over terrain that would be impossible for any wheeled vehicle to handle. 

            How do you design a ‘bot that travels over rough terrain like a mule?  Well, you design it . . . like a mule.  And Boston Dynamics' “Robo-Mule” (later, renamed “BigDog”) was the first in a new line of bio-inspired “walking” robots.  But, again, why a cheetah?  Is it just a cool sounding name or is it the sleek look of the moving animal?  Neither. There’s something special about cheetahs that DARPA wants to capture in robotic performance.

            Robo-Cheetah is being designed to move, quite specifically, like a cheetah.  Unlike Robo-Mule (“BigDog”), Robo-Cheetah is meant to be ultra-speedy and agile, able to “chase and evade” like the actual animal.  Designers are working on getting it to run at cheetah speed, but their ultimate ambition goes much farther than that.  They hope to design a ‘bot that can run faster than any animal on earth -- as fast as 70 mph.

            Robo-Cheetah will have many military applications, including emergency and disaster response.  But DARPA has hinted at performance that might improve on nature.  At least, humans might be able to do things with Robo-Cheetah you’d never try with the real thing – including uses in “advanced agriculture and vehicular travel.”  Just think.  Riding a Robo-Cheetah!

            Of course, the pressure rose with two Robo-Cheetahs in development: The speedy one by Boston Dynamics and the graceful one by MIT.  But, the race got even tighter when another competitor came out of left field -- the Robo-Ostrich.  Ostrich?  What’s an ostrich got to offer in this race?  An ostrich is a bird, and it can’t even fly.  Well, fly it can’t, but maybe it doesn’t need to because the ostrich is the fastest land animal on earth.

            DARPA has funded the joint effort of MIT and the Florida Institute of Human and Machine Cognition (IHMC) in a project to develop a robot that walks and runs.  But the end result of this latest effort will be the first robotic biped in the DARPA arsenal -- a robotic ostrich.

            Robo-Ostrich is designed not just to walk, but to run and run fast.  Although the first full prototype has yet to be designed, the working computer simulation has legs and is hitting speeds of 27 mph.  Impressive, again, because this is about the speed of the fastest human runner in a hundred yard dash.

Robo-Ostrich 

            Robo-Ostrich’s designers are only hoping for a maximum speed of 50 miles an hour – faster than the fastest ostrich -- clocked at 43 mph.  On the other hand, this is a bit slower than the 70 mph Boston Dynamics is hoping for Robo-Cheetah.   But there’s “a whole ‘lot of hoping going on” here.  Robo-Cheetah isn’t off the treadmill, and Robo-Ostrich is a computer simulation.  We’ll just wait and see.

            What’s the secret of Robo-Ostrich’s speed?  Two legs.  What’s so special about a two-legged robot?  Not only is a two-legged robot lighter than a robot with twice the number of legs, but its movements are more flexible allowing it to, among other things, “get through narrower spaces” and maneuver more easily around obstacles.  With such a flexible build, this robot, like other “be-footed” robots, is designed to negotiate rough terrain that would defy a wheeled-vehicle like a jeep.  Even on the most irregular surfaces, the finished ‘bot is expected to run (or walk) at a speed of 10 mph, more than twice as fast as a walking human being.

            Well, with MIT pushing hard to the goal with both their robots, Robo-Cheetah and Robo-Ostrich, Boston Dynamics had to do something.  They announced their plan to take the lead in the race, by unleashing Robo-Cheetah from its treadmill.  They promised their Robo-Cheetah, unteathered, would hit the road in 2013.   And it did, but with a twist.

            In 2013, the cordless “Wildcat” was shown galloping and running (and even running backward) on flat terrain.  But, wait, what happened to Robo-Cheetah?  Why the Wildcat?  Why Robo-Cheetah's little sister?

            To speed up development, Robo-Cheetah was . . . modified.  To get rid of its power cord and, then, to get it off the treadmill and onto the ground, Robo-Cheetah had to lose some of its bulk and weight.   It, also, lost its electric motor and gained an internal combustion (gasoline powered) engine.   Even with these reductions, it lost some of its treadmill speed -- slowing from 28 to about 16 mph. 

            The Wildcat may be slower than the fastest human in a hundred yard dash but, if its chasing you, you’d better get to safety within that hundred yards.  Why?  Because the Wildcat will still be going strong and fast long after you, I, or any human runner, would have worn out and fallen to the ground.  The Wildcat still only performs on flat terrain, but the plan is to have it walking on the same rough ground that its distant cousin the Robo-Mule/BigDog handles with ease.

 



The Wildcat

 

Thursday 6 February 2014

GCLM5444HOxenia

ROBOTS: “Climbing the Walls” – The RiSE Robots

 12 June 2014

            Just when you think you know about all of DARPA’s “legged-robots,” another one pops up.  Or, in this case, climbs up the wall beside you and “surprises” you.

RiSE V1

 

            The RiSE climbing robots, though certainly platforms for the development of future technology, are themselves prototypes designed for field operations and testing.  The RiSE six-legged (“hexapodal”) robots are designed to walk on level ground, but their “claim to fame” is the ability to climb up vertical surfaces.       

 

RiSE V1

            With DARPA funding, “RiSE V-1” was principally designed by Boston Dynamics with “input” from the collaborative consortium including Stanford University, Carnegie Mellon University, U.C. Berkeley, and Lewis & Clark University.  With six legs and two motors, the V1’s vertical climbing abilities were tested on less than smooth surfaces such as a “carpeted wall” and a tree trunk. 

RiSE V2

            The next generation, the RiSE V2, extended the range of “climbable” surfaces.  Like its predecessor, this ‘bot could climb natural, outdoor, vertical surfaces including trees.  But, unlike its predecessor, the V2 could also climb the sides of buildings.  The “body” of the V2 was made larger to hold its power supply, i.e., battery packs.  Because the V2’s had a larger foot mechanism, the body was also made longer with the insertion of “spacers” to allow more clearance for the movement of the ‘bot’s “feet.”

 

RiSE V2 & RiSE V3



            Described as having a “dramatically different gait,” this newest “climber” uses its legs in a different way to, not just to be able to climb poles, but to be a “rapid” pole climber.  Climbing at a bit over 8.5 inches per second, this ‘bot moves up those poles fast. 

RiSE V3

 

            But the continuing development of this ‘bot is aimed at getting it off the byroad and back onto the main drag.  There are two other objectives of the V3’s on-going development.  This ‘bot (1) needs to be able to walk (or run) as fast on flat ground as it does when climbing poles; and (2) need to be able to climb flat walls and other vertical surfaces as well as its “ancestor,” the V2.

            The RiSE robots take there place among the ever-increasing number of DARPA’s  “bioinspired” robotic projects.  The terms “biomimetics” or “biomimicry” have been widely adopted to describe any technology imitating (copied from) nature.  But, in some contexts, biomimicry is more of a necessity than a choice. If you want robots or drone vehicles that work in a particular way, and the only known example of such performance is a biological organism, you’ll either have to imitate the organism or forget the project altogether.

See also: LittleDog Does it First

 

Thursday 12 June 2014

GCLM5444HOxenia

ROBOTS: “Legged” Robots – LittleDog Does It First

5 June 2014

            In 2005, Boston Dynamics unveiled BigDog (also search “Big Dog”) a four-legged (quadrupedal) robot.  The project, funded by DARPA, was intended to develop the robotic equivalent of a pack mule to work directly with soldiers in the field.  As a “legged” robot, BigDog was expected to go where wheeled vehicles couldn’t. 

ROBOTS: Big Dog – Terrestrial Support Robotics

 

 

          Then came “Alpha Dog,” the LS3, an advanced version of BigDog.

 

 

Alpha Dog

ROBOTS: “Alpha Dog” – “Big Dog” Taken to the Next Level

             Boston Dynamics has continued with the on-going development of a faster quadruped, Robo-Cheetah.  Even as Robo-Cheetah took the speed record for “legged” ‘bots, topping 28 mph on an in-lab treadmill, developers remain optimistic that this ‘bot will, someday, reach even higher speeds. 
 

Robo-Cheetah



            While Robo-Cheetah stayed tethered to its treadmill, last October, a slightly smaller and slower spin-off, the Wildcat, jumped off the treadmill.   The Wildcat left its “power cord” behind running at speeds of just over 15 mph.

 

WildCat



            But that’s not all.  A bipedal robo-ostrich, “FastRunner,” has been modeled (computer simulated prototype) as, yet another, more advanced battle ‘bot.  FastRunner’s two legs will allow it to gain more speed and move with more agility than any four-legged robot ever could.

FastRunner / "Robo-Ostrich"

            But let’s take a few steps back.  There’s a smaller less celebrated robot that has had a significant place in the development process of all this robotic technology.  I can’t call this an “unsung” robot, but it’s certainly “less-sung” than the full sized robots we’ve been talking about.  To many, this small robot seems almost like a detail on the R & D trail to the ever-growing family of ever more amazing legged robotic achievements.  But, sometimes, there's more than you'd suspect "in the details."  

LittleDog

            LittleDog was developed by Boston Dynamics with DARPA funding.  Unlike other robotic prototypes, Little Dog was never intended as a stand-alone “field” robot.  LittleDog was, and is, a “testbed.”

            A testbed is a sort of a standard “model” of a device of a certain type -- such as an automobile, airplane, computer, or computer operating system.  This model is used to test new components.  So, let’s say an automobile manufacturer develops the prototype of an innovative new automobile engine or chassis.  The manufacturer’s research division will maintain a sort of “standard” or “model” vehicle into which the newly developed component and be installed and tested. 

            So, LittleDog, “The Legged Locomotion Learning Robot,” is not, and never will be, a robot for use in the field.  Instead, it is a model used to test components being developed for other projects.  And there is more perfected technology stuffed into this small ‘bot than you'd ever guess. 

            Each of LittleDog’s four legs is powered by three electric motors.  At a length of about 12 inches, a width of about 7 inches, and a height of about 5 and a half inches, this small ‘bot can move over obstacles much larger than the length of its legs and body.



LittleDog

            Several separate teams are working at the development of LittleDog’s speed and agility of movement.  All are confident that, if they can make LittleDog do it, the same capacities and abilities can be built into its bigger “field” counterparts.   LittleDog already has such a good “sense of it's surroundings” that it can avoid obstacles that, sometimes, trip-up its comparatively giant “field” counterparts.

            Among other things, LittleDog is trying out new software, which is intended to allow this little ‘bot to read maps and navigate through the corresponding terrain.  Other teams have "taught" this 'bot new walking techniques that allow LittleDog to negotiate obstacles the robot could neither see nor predict.



            While LittleDog may not actually “run with the big dogs,” those bigger dogs can't do anything that LittleDog hasn’t done first.

 

Thursday 5 June 2014

GCLM5444HOxenia