Section 2. Design Alternatives for Miniature Climbing Robots
2.1 Design considerations
Climbing robots, intended for reconnaissance in urban environments, need to be small in size to
avoid detection and have sufficient mobility for travel on exterior and interior surfaces of
buildings. There are few mechanisms suitable for travel on flat inclined surfaces, and suction,
despite limitations, is the inevitable choice for most practical applications. The purpose of our
research has been to design miniature climbing robots based on the suction mechanism with a
high degree of mobility. The design problem is challenging since enhanced mobility necessitates
the use of many actuators but the number of actuators that can be incorporated is limited by the
weight that can be carried by the suction mechanisms. For effective operation in reconnaissance,
the robots also need to be semi-autonomous. The robots should therefore be capable of carrying
multiple sensors, electronic circuits, computational resources, wireless communication devices,
and power supply. More importantly, they should be capable of generating their own suction. All
of this hardware will occupy a significant amount of space and constrain the volume and location
of space available for actuator placement.
A traditional design of our robot would be based on four actuators. Such a design would be
substantially heavy and its weight alone would approach the limits of the suction mechanism.
Furthermore, the space used by four actuators will leave little space for placement of other
hardware. This led us to explore under-actuated designs that incorporate fewer actuators but do
not sacrifice mobility or other functionality. Such designs reduce the weight of the robot for two
good reasons. First, the weight of each actuator constitutes a substantial portion of the total
weight, and second, elimination of an actuator allows us to downsize the others since each
actuator is typically designed to carry the weight of the others. In the next two sections, we
present two under-actuated kinematic designs for climbing robots. The first design uses four
revolute joints and three actuators and results in a lightweight mechanism with good climbing
and walking mobility. In this design, one actuator provides steering capability and the remaining
two actuators propel the robot in a cartwheel style gait. The second design, which incorporates
four revolute joints and a prismatic joint in an under-actuated configuration, is capable of
Figure 1. Cartwheel style gait of revolute hip biped
1
adapting to its environment by using two different forms of locomotion. It is capable of a
crawling stride, which requires less space for maneuvering, and a faster gait where it swings
from foot to foot. Such adaptability is achieved without increasing the number of actuators, and
importantly, the weight of the robot.
2.2 Revolute Joint Biped
We chose a biped configuration for our first robot primarily due to its simplicity, weight savings,
and reduced size. Several joint arrangements were investigated within the biped format but the
fundamental variation considered was in the type of middle joint used. Specifically, three
variations with no middle joint, a revolute middle joint, and a prismatic middle joint, were
considered. After analyzing the mobility of these joint structures walking between inclined
surfaces, the revolute hip format was selected for its superior mobility.
The basic kinematic structure of our first design is schematically shown in Figure 1. The actual
robot is shown in Figure 2. Our robot has five links, links 1 through 5, with one link, link 1 or
link 5, securely fixed to the travelling surface during articulation of the structure. The remaining
four links are driven by three actuators, with one actuator driving two links. Specifically, one
actuator drives joint 1 to steer the robot, a second actuator drives both joints 2 and 3, and a third
actuator drives joint 4. Together, the second and third actuators drive joints 2, 3, and 4 to
generate a cartwheel gait that involves flipping of the robot structure. A belt drive is used to
couple the articulation of joints 2 and 3 and maintain the rotation of joint 3 equal to twice the
rotation of joint 2. In Fig.1, , , , and , denote variables for joints 1 through 4, respectively.
Figure 2. Joint and link structure of revolute hip biped
2
Each joint of the robot, powered individually, would require four actuators. By coupling the
motion of joints 2 and 3, one actuator is eliminated and the weight of the robot significantly
reduced. A reduction in weight is essential for the suction cups to be capable of supporting the
weight of the robot on vertical and other inclined surfaces. A reduction in the number of
actuators also reduces the degrees of freedom and complicates the task of motion planning. This
problem is overcome through improved and extensive task planning.
Prior to introducing coupling between joints 2 and 3, the kinematic structure of the robot was
envisioned for two different modes of locomotion. One of them is “flipping”, discussed above
and shown in Figs.1 and 3, where the robot flips end-over-end. In the other mode, which can be
termed as “crawling”, the robot will lift one foot up, push it ahead, anchor it down on the surface,
lift the trailing foot and retract it. As shown in Fig.3, the crawling stride is similar to that of an
inchworm.
The flipping and crawling strides each possess distinct benefits and limitations. The flipping
stride needs approximately 122.5 cm2 of space to pass through. The crawling stride requires only
57.5 cm2 of space, as shown in Fig.3. The numbers are based on actual dimensions of our first
design. Although the flipping stride requires more space, it generates a faster gait and is suitable
for open environments. For reduced detection and operation in confined environments, the
crawling stride is however more desirable. The flipping stride also creates a longer moment arm
while climbing vertical surfaces and results in a large moment that must be supported by the
suction feet. The crawling stride does not have this limitation but requires four actuators for
maneuvering, rather than three. The weight of the additional actuator is significant and results in
a much larger load on the suction feet. Since the initial operating environment of the robot was
anticipated to be the exterior surface of buildings, low importance was attached to the space
required by gait. The flipping stride was therefore chosen for lighter weight and higher speed.
2.3 Reconfigurable Prismatic Joint Biped
After design and construction of the revolute hip biped, our efforts focused on the design of a
kinematic structure with improved mobility and adaptability. The goal was to design a robot
comparable to the revolute hip biped in size and weight but capable of operation in confined
environments and narrow spaces. Towards this end, we designed a prismatic hip biped capable of
reconfiguring itself to have multiple forms of locomotion. The benefit of the prismatic structure
is that it can walk through confined spaces using a crawling stride while still being able to walk
between surfaces with different inclinations. Our prismatic hip biped can walk through passages
Figure 3. Flipping and crawling modes of locomotion of revolute hip biped.
3
with cross sectional area equal to 52.5 cm2. This is less than the space required by the revolute
hip crawling stride and less than half of the space required by the revolute hip flipping stride,
both shown in Fig.3. In open environments the robot switches to a faster gait, involving turning
on its feet alternately. This is achieved through reconfiguration of the kinematic structure.
As in the case of the revolute hip biped, we encountered stringent space and weight constraints
while designing the prismatic hip biped. We incorporated the concept of under-actuation but
employed an approach different from the one adopted in our first design to improve adaptability.
We designed the robot with three modes of under-actuation and the capability to switch between
them. In this new paradigm, power from a single actuator is distributed amongst a set of joints of
the robot, but not to all of them simultaneously. In each of the three modes of under-actuation, a
particular subset of joints is driven and the remaining joints are locked to prevent rotation. The
locking and release of joints is achieved with pins, which are engaged and disengaged by the
under-actuated mechanism itself. This is very different from the common form of underactuation where passive joints are controlled using dynamic coupling between them and the
active joints. In the sequel we describe the robot mechanism, its three modes of under-actuation
and two locomotion gaits.
Description of Mechanism: The reconfigurable prismatic hip biped is shown in Figure 4. The
Figure 4. The reconfigurable prismatic hip biped.
4
robot uses three actuators to drive five joints. Motors 1 and 3 independently drive joints 1 and 5,
respectively. These joints allow articulation of the feet relative to the legs. Motor 3 is responsible
for controlling joints 2, 3, and 4. Joints 2 and 3 are revolute joints providing steering capability
of the feet relative to the legs. Joint 4 represents the prismatic motion of the legs that allows the
length of the robot to increase and decrease. In Fig.4, variables corresponding to joints 1 through
5 are denoted by , , , d, and , respectively.
A partially exploded view of the prismatic hip biped robot showing the main sub-assemblies is
shown in Fig.5. The robot consists of identical pairs of legs, feet, and ankles. The ankles are
mounted at the ends of the legs and provide support to the feet. The body of the robot consists of
upper and lower halves that support and guide motion of the legs via the indicated outer and
middle roller guides. The upper half of the body also supports motor 2, which is mounted parallel
to the legs in order to reduce the height of the robot. Power is transmitted from motor 2 through
the helical gears to the drive pinion that propels the geared racks 1 and 2. The linear motion of
the geared racks is transmitted to the legs via lock pins, that is, if the lock-pins are engaged with
the racks. The lock pins pass through and are retained within slots (not visible in Fig.5) on the
legs by bearings that are pressed onto the ends of the pins. The pin engages a notch on the rack
Figure 5. Partially exploded view of reconfigurable prismatic hip biped
5
and couples the rack and leg motions. A spring contained within the assembly, not shown, forces
the lock pin to engage the rack. The slots in the legs are designed to allow de-coupling of the legs
and racks. Lock-pin cams contained within the upper and lower halves of the body, shown in
Fig.5, generate the forces necessary to displace the pin for de-coupling. The cams disengage the
lock-pin from the rack as the leg pushes the lock pin bearings into the cam slot. The cam system
is designed such that when the pin disengages from the rack it locks the legs into the body. This
locking effect prevents the leg from moving relative to the body but allows the rack to continue
moving resulting in foot rotation.
First Mode of Under-Actuation (Mode A): Notice that the legs are essentially identical, with
the exception of their lock-pin locations. In the case of leg 1, the lock-pin is mounted at the end
of the leg opposite from the ankle and it enters its cam slot when both legs are extended. In the
case of leg 2, the pin is adjacent to the ankle and enters its cam slot when both legs are
contracted. When the legs and their corresponding racks are locked together, joints 2 and 3 are
prevented from rotating, and rotation of the drive pinion causes prismatic motion of the legs. A
counter-clock-wise (CCW) rotation of the pinion causes the legs to extend while clock-wise
(CW) rotation causes them to contract. This mode of operation, where power from motor 2 is
transferred to joint 4, and joints 2 and 3 are not passive but held fixed, will be referred to as
under-actuation mode-A. If the prismatic motion in mode-A continues beyond a certain range,
both in extension and contraction, one of the lock-pins will enter its cam slot on the body.
Specifically, lock-pin on leg 1 will enter its cam-slot during extension and lock-pin on leg 2 will
enter its cam slot during contraction. As one of the lock-pin enters its cam-slot, it disengages the
corresponding rack from the leg and allows it to move independently. Depending on which lockpin enters the cam-slot, joint 2 or joint 3 is released and actuated, and the robot switches from
mode-A to mode-B or mode-C of under-actuation.
Second Mode of Under-Actuation (Mode B): The robot can switch from mode-A to mode-B of
under-actuation by continuing to expand joint 4 until the lock-pin on leg 1 enters its cam slot and
couples the body and leg. This action simultaneously decouples rack1 from leg 1, which then
extends relative to leg 1 and causes CCW rotation of foot 1 relative to leg 1 and the body about
joint 2. If foot 1 provides the support to the robot, this would result in CW rotation of the robot.
Meanwhile, since the lock-pin on leg 2 still couples the leg and rack motion, leg 2 will continue
to expand and joint 3 will be held fixed. This mode of operation, where power from motor 2 is
transferred to joints 2 and 4, and joint 3 is held fixed, will be referred to as under-actuation
mode-B. The expansion of the body and CCW rotation of foot 1 relative to the body can also
occur with the robot being supported by foot 2. If this is the case, foot 1 can be anchored after it
has turned CCW by some angle. At this time, reversal in the direction of rotation of the drive
pinion will cause foot 1 to rotate CW relative to the body and contraction of the robot body along
joint 4. Since foot 1 now provides support, it will result in CCW rotation of the robot. Therefore,
the second mode of under-actuation can be used for both CW and CCW rotation of the robot
about joint 2.
Third Mode of Under-Actuation (Mode C): If the legs of the robot keep contracting in underactuation mode-A, the lock-pin on leg 2 will enter its cam-slot and couple the body and leg 2.
The leg will be simultaneously decoupled from rack 2, which will then move relative to the leg
and cause CW rotation of foot 2. Since the lock-pin on leg 1 will still couple the leg and its rack,
the robot will continue to contract along joint 4 while joint 2 will be held fixed. This mode of
6
operation, where power from motor 2 is transferred to joints 3 and 4, with joint 2 held fixed, will
be referred to as under-actuation mode-C. In mode C, similar to mode B, the robot can rotate bidirectionally, about joint 3 standing on foot 2.
Locomotion Gait - Crawling: The kinematic structure of the prismatic hip biped is adapted to
perform “crawling” and “pivoting” strides of locomotion. In the crawling stride, the robot will
essentially operate in under-actuation mode-A. Initially perched on one of its feet, say foot 1, the
robot will first lift its body through articulation of its ankle joint, joint 1. The robot will then
expand its body along joint 4 in under-actuation mode-A and articulate joint 1 to lower its body
and bring foot 2 in contact with the walking surface. Once this foot is engaged, the robot will
articulate joint 5 to raise the body and lift foot 1, contract the body along joint 4 in underactuation mode-A, and articulate joint 5 to bring foot 1 back in contact with the walking surface.
The procedure will be repeated to walk along a straight line with the gait resembling the motion
of an inchworm. If the robot needs to steer its direction, it will switch to mode-B or mode-C of
under-actuation, depending on the foot used for steering. After steering, it will revert back to
mode-A for motion along a straight line.
Locomotion Gait - Pivoting: The pivoting stride employs all three modes of under-actuation to
generate a faster gait of locomotion. To explain this gait, we assume the robot to be initially
supported by foot 1 and in mode-A of under-actuation. In this configuration, the robot will lift its
body by articulating joint 1, and then expand its body along joint 4. It will continue to expand till
the lock-pin on leg1 enters its cam slot and switches the mode of under-actuation from A to B.
The robot will now begin to swing on foot 1 in the CW direction. After a desired angle of turn,
the robot will lower its body and anchor foot 2 on the walking surface. It will subsequently raise
its body by articulating joint 5 and drive motor 2 to contract along joint 4. Initially joint 4 will
contract due to contraction of leg 2 alone but after some time rack 1 will engage with leg 1. At
this time the robot will switch from under-actuation mode B to mode A and both legs will
contract along joint 4. As the robot continues to contract along joint 4, the lock-pin on leg 2 will
enter its cam slot and the robot will switch its under-actuation from mode A to mode C. The
robot will subsequently turn on foot 2 in the CCW direction. After a certain angle of turn, the
robot will lower its body and anchor foot 1 on the walking surface. It will subsequently raise its
body by articulating joint 1 and drive motor 2 to expand along joint 4. Initially joint 4 will
expand due to expansion of leg 1 alone but after some time rack 2 will engage with leg 2. At this
time the robot will switch from under-actuation mode C to mode A and both legs will expand
along joint 4. The whole procedure can be repeated to generate alternate turning of the robot
about foot 1 and foot 2 in CW and CCW directions, respectively, and thereby generate the
pivoting mode of locomotion. By keeping the angle of turn to be the same on both feet, the robot
can essentially move in a fixed direction. If the angles of turn are different the robot can
negotiate a curved path.
7