Robotics/Types of Robots/Walkers

“In the early days of air travel, detractors used to argue that, if God had meant us to fly, he would have given us wings. Had he meant us to roll, he might also have given us wheels – but instead we, along with the great preponderance of land animals great and small, have wound up traveling on legs.”

Legs v. Wheels

Most man-made vehicles today travel on wheels and for good reason: wheels are much easier to construct and control. In today’s economy, they also tend to be much cheaper than their legged counterparts. However legs have distinct advantages over wheels. The biggest advantage is in transversability and efficiency. Legged robots have a unique ability to:

• Isolate their body from terrain irregularities
• Avoid undesirable footholds
• Regulate their stability
• Achieve energy efficiency

These advantages are very desirable in modern robotics, and therefore a lot of research is being put into creating robots that can walk. The most challenging task in designing a legged robot is to create a system that can generate the proper gait.

Dynamic v. Static

Locomotion techniques can be divided into two main categories: static and dynamic.

Robots that use static movement are always balanced; that is, their center of gravity is always within their ground contact base. While this technique has been successfully used to create many robots (included wheeled ones), it is more akin to wheeled movement than true dynamic walking and as such retains fewer of the advantages. While more adept at transversing uneven terrain than most wheeled robots, robots that use static walking are very inefficient as power is put into every movement. However, robots that use static walking are much easier to control than their dynamic counterparts and thus often more viable.

Dynamic walking is characterized in that the robot is not always in balance. Many robots that use dynamic walking are continually “falling” and thus much more energy efficient. Dynamic walking requires much more complex control systems in order to not fall. Robots utilizing dynamic walking cannot use the same motions at different speeds to attain different speeds of movement, but must use entirely different motions at different speeds. However, dynamic walking can achieve many more advantages over wheeled locomotion. Dynamic walking is found very abundantly in nature.

A subset of dynamic walking is called passive dynamic movement. Most dynamic walking systems use active control to move the legs to the correct orientations for walking (hence active dynamic walking). Passive dynamic walking is characterized by a system where “gravity and inertia alone generate the locomotion pattern.” Passive dynamic movement can be achieved with maximum efficiency, as the vehicle uses its own forward momentum to propagate its next movement. Very little energy is lost from the system. Most of the concepts of passive dynamic walking and research conducted in the field was done by aeronautical engineer Tad McGeer between 1988 and 1992.

More Than Four

Many different walking robots have been developed that use six or more legs. This is due to the fact that a robot using this many legs can be controlled with static walking techniques rather than dynamic walking. Most of the walking techniques can be demonstrated sufficiently using the six legged model:

Wave Gait

(1)	(2)	(3)	(4)	(5)

1. Six legged robot in neutral position
2. Front pair of legs move forward
3. Second pair of legs move forward
4. Third pair of legs move forward
5. Body follows legs forward

Tripod Gait

(1)	(2)	(3)	(4)

1. Six legged robot in neutral position
2. Alternating legs move forward on either side
3. All other legs move forward
4. Body follows legs forward

Most robots using six or more legs use a variation of one of these two gait models.

Four Legs

A system on four legs is another walking scheme found readily in nature. Four legged robots have the advantage of being statically stable when not moving, but require dynamic walking control. There are many different ways for a four legged robot to walk including alternating pairs and opposite pairs as in six legged robots. However these techniques now cease to be statically stable and thus require dynamic control.

Boston Dynamics has developed a four legged robot for DARPA (Defense Advanced Research Projects Agency) called “Big Dog,” that they claim is “the most advanced quadruped robot on earth.” Big Dog can run at four miles per hour, climb thirty five degree slopes, and carry 340 pounds. But the most impressive feature is its dynamic walking: Big Dog can recover from slipping and even being pushed. Its behavior is such that it approaches the infamous “uncanny valley” (http://en.wikipedia.org/wiki/Uncanny_Valley). The Boston Dynamics website (http://bostondynamics.com/content/sec.php?section=BigDog) features a video demonstration of Big Dog’s abilities.

Three Legs

Three legged robots are not very common, especially since they have no biological counterparts. However, researchers at Virginia Tech’s RoMeLa lab have developed a three legged robot STriDER that uses a “revolutionary” passive dynamic walking technique. STriDER is short for Self-excited Tripedal Dynamic Experimental Robot. STriDER sways until it can lift one leg, and using the other two as an A-frame, swing it in between the other two “stance” legs moving forward at a sixty degree angle. This patent pending “tripedal gait” is extremely energy efficient and requires minimal control. It also allows STriDER to easily change directions by changing the sequence of its steps. A video posted by the development team can be found at http://www.youtube.com/watch?v=7XsaJwKKBYo&feature=related.

Two Legs

Two legged robots have probably seen the most development dollars since humanoid robots have been envisioned since the very beginning of the field. Much of the development in passive dynamic walking has been done in this area. The design of a bipedal passive dynamic walker begins with the concept of a wheel with spokes. If the wheel is divided into sections, and all but two removed, we have what appears to be a set of legs. When the mass is properly distributed, the legs each act as inverted pendulums and the robots “rolls” through its steps.

Further complexity can be added to the model by using knee joints to shorten the legs (allowing one to swing past the other without touching the ground) and ankle joints that can provide a “spring” to the step to add lost energy back into the system. For a more in-depth explanation of passive dynamics walking visit http://www-personal.umich.edu/~artkuo/Passive_Walk/passive_walking.html.

There are several robots that have used these concepts to achieve firsts in the field of robotics. “RunBot,” developed in Germany and Scotland, broke the speed record per size for a robot in April 2006 by walking at 3.5 leg-lengths per second. The Cornell Ranger, while not truly passive, is passive inspired and one of many robots that has more than two legs but is still classified as bipedal. When viewed from the side, Ranger appears to have only two legs, but it actually has four legs. These four legs act in pairs of two, qualifying it as bipedal but providing better lateral stability. On April 3, 2008 Ranger walked 9.07 kilometers without stopping, an unofficial record at that time (it has since been surpassed, according to the Cornell team, by Boston Dynamic’s Big Dog).

One of the most successful companies at building bipedal robots over the years has been Honda. Their most recent model, ASIMO, is one of the few bipedal robots that appears humanoid, can climb stairs, and carries its own power supply. ASIMO can also change its gait in real time using Honda’s i-WALK technology. This allows ASIMO to continuously change speeds and direction. The robot can walk up stairs and run up to four miles per hour.

One Leg

1980 and 1993 there was a lot of research in making one legged robots at the Massachusetts Institute of Technology (MIT). The MIT lab turned out a series of “MIT hoppers” that could balance themselves and traverse a path. The biggest challenge with the hoppers was that they could not stand still; they needed to continue hopping in order to maintain their balance. Researchers were able to build a 3-D One-Leg Hopper that “hopped in place, traveled at a specific rate, followed simple paths, and maintained balance when disturbed.” They also constructed a hopper named Uniroo that used an actuated tail to maintain its balance.

References

1. ^ “McGeer_1992_Chap.pdf (application/pdf Object).” 10 Oct 2008 <http://www-personal.umich.edu/~shc/McGeer_1992_Chap.pdf>.
2. ^ “Gait Generation for Orthogonal Legged Robots.” 13 Oct 2008 <http://ranier.hq.nasa.gov/telerobotics_Page/Technologies/0302.html>.
3. ^ “mcgeer_1990_passive_dynamic_walking.pdf (application/pdf Object).” 13 Oct 2008 <http://ruina.tam.cornell.edu/research/topics/locomotion_and_robotics/history/papers/mcgeer_1990_passive_dynamic_walking.pdf>.
4. ^ “History of Passive Dynamics & Toys.” 13 Oct 2008 <http://ruina.tam.cornell.edu/research/topics/locomotion_and_robotics/history.htm>.
5. ^ “Boston Dynamics: The Leader in Lifelike Human Simulation.” 13 Oct 2008 <http://bostondynamics.com/content/sec.php?section=BigDog>.
6. ^ “Robot Shatters Speed-Walking Record | LiveScience.” 13 Oct 2008 <http://www.livescience.com/technology/060428_speedy_robot.html>.
7. ^ “Cornell Ranger, walking robot.” 13 Oct 2008 <http://ruina.tam.cornell.edu/research/topics/locomotion_and_robotics/papers/CornellRanger/index.html>.
8. ^ “ASIMO - The World's Most Advanced Humanoid Robot.” 13 Oct 2008 <http://asimo.honda.com/InsideASIMO.aspx>.
9. ^ “MIT Leg Lab Robots.” 13 Oct 2008 <http://www.ai.mit.edu/projects/leglab/robots/robots-main.html>.