Gait
Adaptation to Environmental Disturbances
Abstract
Natural environments and intrinsic robot dynamics
can produce instability in walking-robot gaits. In such cases, the gait should
be modified to enhance the robot’s stability. This work proposes a novel
gait-adaptation method based on the maximization of a dynamic energy stability
margin. This method enables walking-machine gaits to adapt to internal and
environmental perturbations as well as to the slope of the terrain by finding
the gait parameters that maximize robot stability. The adaptation method also
gives mathematical insight into the natural gait adaptation carried out by
humans and animals to balance external forces or the effect of sloping terrain.
Experiments with the SILO4 robot are presented that show how robot stability is
enhanced when the proposed approach is used for different external forces and
sloping terrains.
In walking-robot locomotion, stability must be carefully controlled during a given gait cycle to prevent the machine from tumbling down. Walking robots designed for field and service applications usually perform statically stable gaits. Such gaits are defined to optimise the Static Stability Margin. However, walking robots used for field and service applications are heavy-limbed machines that carry a robot manipulator or manipulators to perform their tasks, like the robot shown in Figure 1. Dynamic effects caused by the motion of heavy limbs and robot manipulators perturb robot stability due to the lack of dynamic information in the computation of the Static Stability Margin. The resulting problem in walking-robot stability control has led to the definition of several dynamic stability margins that allow for robot dynamics.
|
Figure 1: DYLEMA: a walking
platform for antipersonnel mine detection |
Recent research on the quantitative
and qualitative classification of these stability margins has shown that the
most accurate stability margin when robot dynamics are significant is the
Normalized Dynamic Energy Stability Margin, NDESM. However, even when the
suitable stability margin is used to control the robot, the robot will still
tumble down when confronted with any perturbing effect. The reason for this
contradiction is the use of robot gaits that have been designed to optimize a
static-stability margin. Although an accurate dynamic-stability margin measures
robot stability, unstable situations are merely observed, not avoided,
unless the gait pattern is modified.
Therefore, robot gaits should be modified based on the more adequate NDESM to
improve gait control.
A
clarifying example of this problem can be found in nature. Humans and animals
adapt their gait to terrain inclination and external forces. Figure 2 shows how
human gait parameters change when a person walks uphill or counteracts an
external force. The effect of an inclined terrain or a counteracting force on
gait is to reduce both the leg stroke (d) and the center-of-gravity (CG) height (h). The leg-stroke reduction implies the
modification of footholds in the gait cycle.
Another effect is the modification
of the horizontal position of the CG. These three parameters are modified to
enhance stability. Therefore, the computation of a stability margin that
considers the destabilizing effect of both ground inclination and external
forces -such as the Force-Angle Stability Margin and the
Normalized Dynamic Energy Stability
Margin - is not enough to enhance stability. The gait parameters must also be
recomputed to achieve efficient motion control.
This work proposes a method for gait
optimization based on the NDESM. The method maximizes the stability margin
during the transfer and support phases. As the SILO6 platform, shown in Figure
1 is not ready, experiments using the SILO4 robot have been performed to show the improvement
of the adaptive gait-optimization method when external forces are applied and
when the robot walks on sloping terrain.
|
Figure 2: CG height (h) and leg stroke (d) of a person
walking on a) flat terrain; b) inclined terrain; c) flat terrain under
external forces |
The gait-adaptation process was performed for the SILO4 quadruped robot. Figure 3 shows the gait parameters obtained in the optimization procedure for different terrain inclinations and external forces.
(b)
Figure 3: Gait parameters for robot adaptation to (a) sloping ground, (b) external forces
Experiments have been carried out to
show the improvement achieved on the walking gait. In the experiment the SILO4
robot walks on flat ground and it is affected by a 50-N external force, that is
achieved by attaching the robot to a load. The NDESM was computed while the
robot was carrying the load using a two-phase discontinuous gait. The robot
tumbles down when no gait adaptation method is used (see video1 10 MB MPEG). Then, the experiment
is repeated using the gait adaptation approach (see video2 6 MB MPEG), using the gait patterns listed in Table I for an external force of
50N, which reduces CG height and leg stroke to maximize the stability margin. In
this experiment the robot did not tumble down, therefore the resulting gait
increases its robustness to external disturbances. By comparison of both videos
one can observe how much the height of the gait pattern is reduced as well as
the leg stroke to increase robot stability when the gait adaptation algorithm
is used.
Video 1: SILO4 carrying a load using a Video 2: SILO4 carrying a load using gait adaptation indoors
2-phase discontinuous gait indoors
Video 3: SILO4, 2-phase gait natural environment Video 4: SILO4, gait adaptation, natural environment