Sandin P.E. Robot mechanisms and mechanical devices illustrated

Подождите немного. Документ загружается.

194 Chapter 6 Steering History

vehicles use this method to steer, while older tracked vehicles would

brake a track on one side, slowing down only that track, which turned

the vehicle.

As discussed in the chapter on wheeled vehicles, this is also the steer-

ing method used on some four-wheel loaders like the well-known

Bobcat. One motor drives the two wheels on one side of the vehicle, the

other drives the two wheels on the other side. This steering method is so

effective and robust that it is used on a large percentage of four-, six-, and

even eight-wheeled robots, and nearly all modern tracked vehicles

whether autonomous or not. This steering method produces a lot of skid-

Figure 6-5a Differential

steering

Figure 6-5b

Chapter 6 Steering History 195

ding of the wheels or tracks. This is where the name “skid steer” comes

from.

The fact that the wheels or tracks skid means this system is wasting

energy wearing off the tires or track pads, and this makes skid steering an

inefficient design. Placing the wheels close together or making the tracks

shorter reduces this skidding at the cost of fore/aft stability. Six-wheeled

skid-steering vehicles can place the center set of wheels slightly below

the front and back set, reducing skidding at the cost of adding wobbling.

Several all-terrain vehicle manufacturers have made six-wheeled vehi-

cles with this very slight offset, and the concept can be applied to indoor

hard-surface robots also. Eight-wheeled robots can benefit from lower-

ing the center two sets of wheels, reducing wobbling somewhat.

The single wheel drive/steer module discussed earlier and shown on a

tricycle in Figure 6-6 can be applied to many layouts, and is, in general,

an effective mechanism. One drawback is some inherent complexity

with powering the wheel through the turning mechanism. This is usually

accomplished by putting the drive motor, with a gearbox, inside the

wheel. Using this layout, the power to the drive motor is only a couple

wires and signal lines from whatever sensors are in the drive wheel.

These wires must go through the steering mechanism, which is easier

than passing power mechanically through this joint. In some motor-in-

wheel layouts, particularly the syncro-drive discussed next, the steering

Figure 6-6 Drive/steer module

on tricycle

196 Chapter 6 Steering History

mechanism must be able to rotate the drive wheel in either direction as

much as is needed. This requires an electrical slip ring in the steering

joint. Slip rings, also called rotary joints, are manufactured in both stan-

dard sizes or custom layouts.

One type of mechanical solution to the problem of powering the

wheel in a drive/steer module has been done with great success on sev-

eral sophisticated research robots and is commonly called a syncro-

drive. A syncro-drive (Figure 6-7) normally uses three or four wheels.

All are driven and steered in unison, synchronously. This allows fully

holonomic steering (the ability to head in any direction without first

requiring moving forward). As can be seen in the sketch, the drive

motor is directly above the wheel. An axle goes down through the cen-

ter of the steering shaft and is coupled to the wheel through a right angle

gearbox.

This layout is probably the best to use if relying heavily on dead reck-

oning because it produces little rotational error. Although the dominant

dead-reckoning error is usually produced by things in the environment,

this system theoretically has the least internal error. The four-wheeled

layout is not well suited for anything but flat terrain unless at least one

wheel module is made vertically compliant. This is possible, but would

produce the complicated mechanism shown in Figure 6-8.

Figure 6-7 Synchronous drive

Chapter 6 Steering History 197

All-terrain cycles (ATCs), when they were legal, ran power through a

differential to the two rear wheels, and steered with the front wheel in a

standard tricycle layout. ATCs clearly pointed out the big weakness of

this layout, the tendency to fall diagonally to one side of the front wheel

in a tight turn. Mobility was moderately good with a human driver, but

was not inherently so.

Quads are the answer to the stability problems of ATCs. Four wheels

make them much more stable, and many are produced with four wheel

drive, enhancing their mobility greatly although they cannot turn in

place. They are, of course, designed to be controlled by humans, who

can foresee obstacles and figure out how to maneuver around them. If a

mobility system in their size range is needed, they may be a good place

to start. They are mass-produced, their price is low, and they are a

mature product. Quads are manufactured by a number of companies

and are available in many size ranges offering many different mobility

capabilities.

As the number of wheels goes up, so does the variety of steering

methods. Most are based on variations of the types already mentioned,

but one is quite different. In Figure 4-30 (Chapter Four), the vehicle is

divided into 2 sections connected by a vertical axis joint. This layout is

common on large industrial front-end loaders and provides very good

steering ability even though it cannot turn in place. The layout also

Figure 6-8 Drive/steer module

with vertical compliance

198 Chapter 6 Steering History

forces the sections to be rather unusually shaped to allow for tighter turn-

ing. Power is transferred to the wheels from a single motor and differen-

tials in the industrial version, but mobility would be increased if each

wheel had its own motor.

Chapter 7 Walkers

This page intentionally left blank.

here are no multi-cell animals that use any form of continuously

rolling mechanism for propulsion. Every single land animal uses

jointed limbs or squirms for locomotion. Walking must be the best way

to move then, right? Why aren’t there more walking robots? It turns out

that making a walking robot is far more difficult than making a wheeled

or tracked one. Even the most basic walker requires more actuators,

more degrees of freedom, and more moving parts.

Stability is a major concern in walking robots, because they tend to be

tall and top heavy. Some types of leg geometries and walking gaits pre-

vent the robot from falling over no matter where in the gait the robot

stops. They are statically stable. Other geometries are called “dynami-

cally stable.” They fall over if they stop at the wrong point in a step.

People are dynamically stable.

An example of a dynamically-stable walker in nature is, in fact, any

two-legged animal. They must get their feet in the right place when they

want to stop walking to prevent tipping over. Two-legged dinosaurs,

humans, and birds are remarkably capable two legged walkers, but any

child that has played Red-light/Green-light or Freeze Tag has figured out

that it is quite difficult to stop mid-stride without falling over. For this

reason, two legged walking robots, whether anthropomorphic (human-

like) or birdlike (the knee bends the other way), are rather complicated

devices requiring sensors that can detect if the robot is tipping over, and

then calculate where to put a foot to stop it from falling.

Some animals with more than two legs are also dynamically stable

during certain gait types. Horses are a good example. The only time

they are statically stable is when they are standing absolutely still. All

gaits they use for locomotion are dynamically stable. When they want to

stop, they must plan where to put each foot to prevent falling over.

When a horse’s shoe needs to be lifted off the ground, it is a great effort

for the horse to reposition itself to remain stable on three hooves, even

though it is already standing still. Cats, on the other hand, can walk with

a gait that allows them to stop at any point without tipping over. They do

not need to plan in advance of stopping. This is called statically-stable

independent leg walking. Elephants are known to use this technique

201

202 Chapter 7 Walkers

when crossing streams or difficult terrain. They stand on three legs

while the forth leg is moved around until it finds, by feel, a suitable

place to set down.

These examples demonstrate that four-legged walkers can have

geometries that are either dynamically or statically stable or both.

Animals have highly developed sensors, a highly evolved brain, and fan-

tastically high power-density muscles, that allow this variety of motion

control. Practical walking robots, because of the limitations of sensors,

processors, and fast acting powerful actuators, usually end up being stat-

ically stable with two to eight legs.

The design of dynamically-stable, walking mobile robots requires an

extensive knowledge of fairly complicated sensors, balance, high-level

math, fast-acting actuators, kinematics, and dynamics. This is all beyond

the scope of this book. The rest of this chapter will focus on the second

major category, statically-stable walkers.

LEG ACTUATORS

First, let’s look and leg geometries and actuation methods. There are

three major techniques for moving legs on a mobile robot.

• Linear actuators (hydraulic, pneumatic, or electric)

• Direct-drive rotary

• Cable driven

Hydraulics is not covered in this book, but linear motion can be done

effectively using two other methods, pneumatic and electric. Pneumatic

cylinders come in practically any imaginable size and have been used in

many walking robot research projects. They have higher power density

than electric linear actuators, but the problem with pneumatics is that the

compressed air tank takes up a large volume.

Linear actuators have the advantage that they can be used directly as

the leg itself. The body of the actuator is mounted to the robot’s chassis

or another actuator, and the end of the extending segment has a foot

attached to it. This concept has been used to make robots that use

Cartesian and cylindrical coordinate walkers. These layouts are not cov-

ered in this book, but the reader is urged to investigate them since they

can simplify the development of the control code of the robot.

Chapter 7 Walkers 203

Direct-drive rotary actuators usually have to be custom designed to

get torque outputs high enough to rotate the walker’s joints. They have

low power density and usually make the walker’s joints look unnaturally

large. They are very easy to control accurately and facilitate a modular

design since the actuator can be thought of conceptually and physically

as the complete joint. This is not true of either linear actuated or cable

driven joints.

Cable-driven joints have the advantage that the actuators can be

located in the body of the robot. This makes the limbs lighter and

smaller. In applications where the leg is very long or thin, this is critical.

They are somewhat easy to implement, but can be tricky to properly ten-

sion to get good results. Cable management is a big job and can consume

many hours of debug time.

LEG GEOMETRIES

Walking robots use legs with from one to four degrees of freedom

(DOF). There are so many varieties of layouts only the basic designs are

discussed. It is hoped the designer will use these as a starting point from

which to design the geometry and actuation method that best suits the

application.

The simplest leg has a single joint at the hip that allows it to swing up

and down (Figure 7-1). This leg is used on frame walkers and can be

actuated easily by either a linear or rotary actuator. Since the joint is

already near the body, using a cable drive is unnecessary. Notice that all

the legs shown in the following figures have ball shaped feet. This is nec-

essary because the orientation of the foot is not controlled and the ball

gives the same contact surface no matter what orientation it is in. A sec-

ond method to surmount adding orientation controlled feet is to mount

the foot on the end of the leg with a passive ball joint.

The following four figures show two-DOF legs with the different

actuation methods. These figures demonstrate the different attributes of

the actuation method. Figure 7-2 shows that linear actuators make the

legs much wider in one dimension but are the strongest of the three.

Figure 7-3 shows a mechanism that keeps the second leg segment verti-

cal as it is raised and lowered. The actuator can be replaced with a pas-

sive link, making this a one-DOF leg whose second segment doesn’t

swing out as much as the leg shown in Figure 7-2.