Alternative Gaits For Multiped Robots With Leg Failures To Retain Maneuverability
Alternative Gaits For Multiped Robots With Leg Failures To Retain Maneuverability
Alternative Gaits For Multiped Robots With Leg Failures To Retain Maneuverability
International Journal of Advanced Robotic Systems, Vol. 7, No. 4 (2010) 31
ISSN 1729‐8806, pp. 33‐40
International Journal of Advanced Robotic Systems, Vol. 7, No. 4 (2010)
or joint loses most of its resistance to external load, section we proposed that by introducing the use of
regards the damaged leg as a free‐rotating hanging limb. removable sliding legs, it is possible to recover some
This paper focuses on “Severed Leg” scheme and efficiency and stability.
“Sliding Leg” approach to maintaining the efficiency and The proposed idea for a multiped robot having
stability of robot. Because, damaged leg can produce removable sliding legs is shown in Fig. 1. In this gait plan,
some unwanted side effects to other legs and the robot first joint of each leg is mounted with sliding mechanism
system. Thus, it can be a good practice to discard the on which the leg can freely moves or even detaches itself
damaged leg and slide to better position of other legs to by the command of the operator. Now take (6‐2‐L3R3)‐
recover the robot’s stability and efficiency. This paper type as an example. When legs L3 and R3 failed, they were
assumes that the robot center of gravity and geometric severed and leg L2 taken position of L3, similarly R2 also
center of gravity overlapped, and each leg are able to moved to replace R3 by the command of operator. We
release freely through the control away from the body, named this strategy Fixed Position Adjustment (FP). In
and the weight of the leg is negligible. In addition, this this way, we can convert a low‐efficiency low‐stability (6‐
paper assumes that robot will walk on flat terrain (Pal et 2‐L3R3)‐type to a much satisfactory (6‐2‐L2R2)‐type gait
al., 1991). This paper uses straight forward gaits as an plan. Moreover, we can increase stability and efficiency of
example. Other kind of gaits, such as Crab Gaits or (6‐2‐L1 R3)‐type by similar conversion. Note that the
Turning Gait may be applied as an extension (Wilson, mechanism proposed here for severed and sliding legs
1966). can be remotely controlled by the operator.
This paper assumes to simplify the problem that, after a
fault Robot system can automatically unloaded the leg by
the command of operator, and also the weight of the leg
ignored. But in real application, the severed leg scheme
must be integrated with software and hardware to make
a detailed analysis and design. Moreover, it should
consider the weight of severed leg to measure the center
of gravity of robot system.
Guides
2. The Multiped Robot model
Sliding Device
Leg can detach by the
We considered hexapod and octopod robot model as a command of operator
Multiped robot model in this paper. A three dimensional
coordinate system is used in which the origin is placed at Fig. 1. Robot gait with removable sliding legs. Each leg
the Center of Gravity of the robot. Moreover, leg modules attached with sliding device and mounted on a guide.
are designed in a way that it can discard and slide when The leg can freely moves or even detaches itself by the
needed, as described in (Tsai, C. H. 2008). Each leg has command of the operator. 1st step: Leg is removed, 2nd
three degrees‐of‐freedom, and it swings backward (as in a step: Middle leg slide into the removed leg (Fixed
propulsive stroke) and forward (as in a return stroke) Position Adjustment)
with the same velocity. In this paper, we assumed that
maximum stride length of each leg be 2S, that means a leg When the position of the leg started to move and
can move ± S from its neutral position. remaining leg could not provide stable support, then we
must use a special transfer method to achieve the stable
2.1. The use of removable sliding legs condition. Let’s examine the (6‐2‐L3 R3)–type case, legs L2
Yang (2005b) mentioned that each joints of robot’s leg and R2 are the ones to move their positions. However as
system was assumed to be able to lock individually. Yang any of these legs lifts above the ground, the remaining
further suggested that it was advantageous to lock the legs fail to provide a usable support polygon. In this case
joint associated with a damaged motor. The leg module changing the positions of these legs has to be
with a locked joint was then only be used to support accompanied by a proper movement of the body. For
rather than to propel, and the gait sequence of the robot example firstly let L1 and R1 swing backward a distance S,
had to be re‐arranged accordingly and continue to move and secondly make L2 and R2 also move in the same
forward. But locking mechanism cannot always provide direction an amount of S, then swing L1 and R1 back to
support. For example, when breakdown occurs at second original stances. We need to repeat these steps until L2
or third joint in a leg, then it is highly possible that leg and R2 move to desired fixed positions. In general, (6‐2‐
can no longer provide support. In such a case, we n) and (8‐4‐n) types (only four legs left) needs this kind of
proposed that the leg has better to be removed otherwise accommodation. Other possible design to facilitate the
it will interfere the proper movement of neighboring legs. transportation of the legs are included in (Tsai, C. H. 2008).
When a robot suffers from a leg loss, its gait efficiency as
well as stability reduced significantly. However, in this
32
Kazi Mostafa, Chong‐Shuen Tsai and Innchyn Her: Alternative Gaits for Multiped Robots with Leg Failures To Retain Maneuverability
(a)
Fig. 2. Hexapod and Octapod Robot Schematics.
Maximum stride length is 2S; SL+ and SL‐ are the front
and rear axial stability limit respectively.
3.1. Common Leg
Most insects can walk with the tripod gait, which is a fast
and statistically balanced gait (SBG) for Hexapods. (b)
However, when one or more legs are missing, regular Fig. 3. The 6‐1‐R2 case, as illustrated by robot schematics
tripod gait is no longer possible. To get around this, a (a), and by the Enhanced Gait Chart (EGC) of the robot
“common leg” needs to be shared in two tripod groups. (b). A hexapod robot which has no R2 leg. In 2(a), hollow
We will explain more about “common leg” with an arrow indicates half step & solid arrow indicates full step
example in the following section (2S). Dash lines are indicate the support polygon. In each
interval the robot travels a distance S.
33
International Journal of Advanced Robotic Systems, Vol. 7, No. 4 (2010)
34
Kazi Mostafa, Chong‐Shuen Tsai and Innchyn Her: Alternative Gaits for Multiped Robots with Leg Failures To Retain Maneuverability
35
International Journal of Advanced Robotic Systems, Vol. 7, No. 4 (2010)
(dmin) is thus increased from 1.25S to 1.51S. This is now
the optimal configuration for [2|4] category.
Fig. 5. Best combination for [3|3] type gait sequence
Fig. 7. Best combination for [2|4] type
6. Sliding gait in Circular Hexapod Robot
From the above study, we can see that five legs [2 | 3]
Fig. 6. Best combination for [2|3] type gait sequence
type is the best one. According to Inagaki (1997) analysis,
the five legs when arranged in a regular pentagon, allows
For (8‐3‐L2 L4 R2)‐type, there is a triangle supported by a
both groups to get a better stability. But it’s purely based
ʺpeakʺ (indicated by dashed line circle) can be
on the assumption of the model analysis, did not talk
simultaneously affect the stability of front and rear side,
about the conversion scheme of physical robot.
but will not impact on the overall stability. Before
adjustment (Fig. 6a) two vertices of the support polygon
are L3 and R3, and L3 is the common leg. If we adjust L3 to
increase the (SL‐) of the (L1, L3, R3) group, conversely the
(SL‐) of the (L3, R1, R4) group is reduced. Thus dmin cannot
be increased as a whole. However, (8‐3‐L2L4R2) type can
be transformed into (8‐3‐L1L4R3) type of the same [2|3]
category. The common leg is now R1, as shown in Fig. 6b,
and the other vertices become L2 and L3. In this case R3
can be transported further backward, and the (SL+) for
the (L3, R1, R3) group is increased from 1.25S to 1.5S. Then
L3 is to be transported downward by 0.03S so that (SL+)
and (SL‐) become equally 1.51S. As for the (L2, R1, R4)
group, its (SL+)=2.64S, and (SL‐)=1.52S before adjustment.
If we want (SL+) and (SL‐) of this group to be as close as
possible, without introducing interference, L2 needs to be
moved downward by 0.53S, and (SL+) and (SL‐) become
2.41S and 1.76S, respectively. This is now the optimal
configuration for the [2|3] category.
As shown in Fig. 7, without leg transportation, the (8‐2‐
L1L4) type is the best in [2|4] category. The rules of
making adjustment are similar to those of the (L3, R1, R3)
group in the (8‐3‐L1L4R3) type. The R3 leg in (L3, R1, R3)
support polygon is first moved downward by 0.5S,
making (SL+) increase from 1.25S to 1.5S, then L3 is move
downward by 0.03S to equalize the two SL values to
1.51S. Since the other group (L2, R2, R4) is of opposite
shape, so it is adjusted reversely. The minimal SL value
Fig. 8. Optimal configuration of Circular Hexapod Robot
36
Kazi Mostafa, Chong‐Shuen Tsai and Innchyn Her: Alternative Gaits for Multiped Robots with Leg Failures To Retain Maneuverability
As shown in Figure 8, is a circular hexapod Robot, so type if proposed Fixed Position Adjustment (FP) used.
named because linking its legs constitutes a regular Moreover, after applying FP, the Robot gait can be
hexagon. A special characteristic of this robot is that it adjusted more by introducing Non Fixed Position
moves freely in all directions. Failure of any one leg will Adjustment (NFP). By following this procedure, a best
produce the critical stability Configuration (Figure 8b). By leg configuration can be achieved which can maximize
using circular guide robot and stepless adjustment, we the stability and efficiency.
converted it pentagon (Figure 8c), or a better arrangement
(Figure 8d), to avoid the occurrence of critical stability 8. Conclusions
condition.
In real application, the construction and control method
7. Procedure to resume its task after leg failure of multi legged robot is complicated. But if a leg failure
occurred, then it needs some correction in gait to
This paper also proposes a procedure for a multi legged maintain efficiency. In some mission such as: planetary
robot to resume its task after a serious leg failure. As exploration, disaster response or antiterrorism mission; it
shown in Fig. 9, whether the damaged legs can provide is impossible to repair immediately after damage
support has to be evaluated immediately after a occurred. Then robot needs some alternative strategies to
breakdown is reported. The probability for a leg not complete its mission. In this paper we propose a
being able to support is estimated as 67% (Fault at 2nd & removable sliding leg approach to solve this problem. We
3rd Joint). So, damaged leg has to be severed. After leg focused on “Severed Leg” scheme and “Sliding Leg”
severance the resulting leg configuration is compared approach to maintaining the efficiency and stability of
with the optimal configuration of the category. robot. Because, damaged leg can produce some
unwanted side effects to other legs and the robot system.
Thus, it can be a good practice to discard the damaged leg
and slide to better position of other legs by the command
of operator to get optimum alternative gait configuration.
For gait adjustment, we proposed Fixed Position
Adjustment (FP) and Non Fixed Position Adjustment
(NFP). By taking the proposed Fixed Position Adjustment
(FP) and Non Fixed Position Adjustment (NFP), the
multiped Robot can overcome any fault event and
maintain stability and efficiency. Moreover, we studied
the use of removable sliding leg approach on hexapod
and octopod robots with leg failures. Based on leg
sequence, stride length, longitudinal stability and
efficiency, alternative gaits are evaluated. In addition, we
recommended tables for different gait sequence with
progressive efficiency and a procedure for a multi‐legged
robot to complete its mission after serious leg failure
happened. These tables and procedures can provide
options for alternative gait and information about certain
damaged leg. It is therefore advisable for a robot designer
to have these preferred gaits programmed into the robot’s
memory. Thus in an adverse condition, it will know
which alternative gait pattern it should turn to.
9. References
Ahmed, M.; Khan, M. R.; Billah, M. M. & Farhana, S.
(2010). Walking Hexapod Robot in Disaster Recovery:
Developing Algorithm for Terrain Negotiation and
Navigation, New Advanced Technologies, Edited by
Fig. 9. Procedure for a Multi legged robot to complete its Aleksandar Lazinica, pp. 341‐350, Publisher INTECH,
mission after serious leg failure ISBN 978‐953‐307‐067‐4, India
Baudoin, Y.; Acheroy, M; Piette, M. & Salmon, J.P. (1999).
Table 2 showed the probability of the occurrence of an Humanitarian demining and robotics, Humanitarian
optimal type. For instance, for (8‐3‐n) category, the demining and robotics, Mine Action Inform Center J 3
occurrence for the best type is only 17%. However, non‐ (2).
optimal types can always be transformed to an optimal
37
International Journal of Advanced Robotic Systems, Vol. 7, No. 4 (2010)
Boissonnat, J. D.; Devillers, O.; Donati, L. & Preparata, F. Learning & Data Mining for Robotics, pp. 39‐49, Bled,
P. (1992). Motion planning for spider robots, IEEE Slovenia, September 2009
International Conference on Robotics and Song, S. M. & Choi, B. S. (1989). A study on continuous
Automation, pp. 2321‐2326, Nice, France, May 1992 follow‐the‐leader (FTL) gaits: an effective walking
Clark, J. E. (2004). Design, Simulation, and stability of a algorithm over rough terrain, Mathematical
hexapedal running robot, PhD thesis, Stanford Biosciences, Vol. 97, No. 2, pp. 199‐233.
University, ISBN: 0‐496‐04396‐X, Stanford Spenneberg, D.; McCullough, K. & Kirchner, F. (2004).
University Stanford, CA, USA Stability of walking in a multilegged robot suffering
Englisg, J. D. & Maciejewski, A. A. (2004). Fault tolerence leg loss, IEEE Conference on Robotics and Automation,
for kinematically redundant manipulators : Vol. 3, pp. 2159‐2164
anticipating free‐swinging joint failures, IEEE Tsai, C. H. (2008). “Gait Algorithm of Multilegged
Transactions on Robotics and Automation, Volume 14, Biomimetic Robots with Leg Failures,” M.S. Thesis,
No. 4, pp. 566‐575, ISSN: 1042‐296X National Sun Yat‐sen University, Kaohsiung, Taiwan,
Englisg, J. D. & Maciejewski, A. A. (2000). Measuring and R.O.C., (in Chinese)
reducing the euclidean‐space effects of robotic joint Wang, P. & Sun, L. (2006). The Stability Analysis for
failures, IEEE Transactions on Robotics and Automation, Quadruped Bionic Robot, International Conference on
Volume 16, No. 1, pp. 20‐28, ISSN: 1042‐296X Intelligent Robots and Systems, October 9‐ 15, 2006,
Guangtao, J .; Haojun, J.; Jinsong, W. & Tiemin, L. (2003). Beijing, China
Petri‐net‐based coordination motion control for Wilson, D. M. (1966). Insect Walking, Annual Review of
legged robot, IEEE transactions on systems, man, and Entomology, Vol. 11, (Jan. 1966), pp. 103‐122, doi:
cybernetics, Vol. 1, pp. 581‐586, ISSN: 1062‐922X 10.1146/annurev.en.11.010166.000535
Habib (2007). Humanitarian Demining: Reality and the Yang, J. M. & Kim, J. H. (1998). Fault‐tolerant locomotion
Challenge of Technology – The state of the Arts, of the hexapod robot, IEEE transactions on systems,
International Journal of Advanced Robotic Systems, Vol. 4, man, and cybernetics, Vol. 28, No. 1, pp. 109‐116,
No. 2, pp. 151‐172, ISSN 1729‐8806 Yang, J. M. (2005a). Tripod gaits for fault tolerance of
Inagaki, K. (1997). Gait study for hexapod walking with hexapod walking machines with a locked joint failure,
disabled leg, International Conference on Intelligent Robotics and Autonomous Systems, Vol. 52, No. 2‐3,pp.
Robots and Systems (IROSʹ97), Volume 1, pp. 408‐413. 180‐189, Doi: 10.1016/j.robot.2005.04.002
Liang, W.; Yuan‐zong, L.; Tie, W. & Zhi‐guo, D. (2010). Yang, J. M. (2005b). Gait Synthesis for hexapod robots
Physical Design on a Robot for Under‐mine Detect with a locked joint failure, Robotica, Vol. 23, No. 6,
and Rescue, Coal Mine Machinery, Vol.31, No. 2, (November 2005), pp. 701‐708, ISSN:0263‐5747
China Academic Journals Electronic Publishing House,
Web: http://www.cnki.net/
McGhee, R. B. & Iswandhi, G. I. (1979). Adaptive
locomotion of a multi‐legged robot over rough terrain,
IEEE transactions on systems, man, and cybernetics,
Volume 9, No. 4, pp. 176‐182.
Mori, Y.; Takayama, K.; Adachi, T.; Omote, S. &
Nakamura, T. (2005). Feasibility study on an
excavation‐type demining robot, Autonomous Robots,
Vo l. 18, No. 3, pp. 263–274, ISSN 0929‐5593 (Print)
1573‐7527 (Online)
Pal, P. K. & Jayarajan, K. (1991). Generation of free gait ‐ a
graph search approach, IEEE Transactions on Robotics
and Automation, Volume 7, No. 3, pp. 299‐305,
ISSN: 1042‐296X
Prahacs, C.; Saunders, A.; Smith, M. K.; McMordie, D. &
Buehler, M. (2005). Towards legged amphibious
mobile robotics, Journal of Engineering Design and
Innovation (online), Vol. 1, part. 01P3
Rizo, J.; Coronado, J.; Campo, C. ; Forero, A.; Otalora, C.;
Devy, M.; et al. (2003). URSULA: robotic demining
system. In: Proceedings of the 11th international
conference on advanced robotics; pp. 538–43.
Saitta, L.; Giordana, A. & Galassi, U. (2009). Robotics in
Planetary Exploration, International workshop on
38