Optimization of Artificial Muscle
Placements for a Humanoid Bipedal
Robot
Connor Morrow
(B) , Benjamin Bolen , and Alexander J. Hunt
Department of Mechanical and Materials Engineering, Portland State University,
Portland, OR 97207, USA
comorrow@pdx.edu
Abstract. This work demonstrates an algorithm that is able to compute
optimal placement for braided pneumatic actuators on a bipedal robot in
order to emulate the biology of human legs. The algorithm calculates the
torque that muscles are able to generate about a series of joints (back,
hip, knee, ankle, subtalar, and metatarsophalangeal) in a human model.
It then compares these torques to the torque that is achievable by a
reduced number of pneumatic muscles actuating a bipedal robot model
and optimizes the results to reduce the error between the robot and
human model. The algorithm successfully finds new muscle placements
that will be used in physical testing to verify that the torque output is
correct and matches similarly to human capabilities. The algorithm was
performed for three muscles about the back (lumbrosacral) joint: erector
spinae, internal oblique, and external oblique. It generates placements
capable of producing torque profiles that are more biologically realistic than the previously hand placed locations. Currently, the algorithm
is not prevented from placing muscle paths that intersect the physical
structure. This work will enable the development of controllable, physical models that more accurately capture force and torque capabilities of
human muscles. Such physical models will enable more complete testing
of how the nervous system performs effective control of over-actuated
muscle systems, and if such systems have advantages for robotic applications.
Keywords: Biomimetic robotics · Optimization algorithm ·
Pneumatic artificial muscle · Braided pneumatic actuator · Opensim ·
Matlab
1 Introduction
Biomimetic robots are an important topic of research to the robotics community,
due to their proposed ability to navigate through environments that are difficult for traditional robots. Robots have trouble navigating through spaces that
contain uneven surfaces or stairs, as well as areas in which there is a transition
c
Springer Nature Switzerland AG 2020
V. Vouloutsi et al. (Eds.): Living Machines 2020, LNAI 12413, pp. 257–269, 2020.
https://doi.org/10.1007/978-3-030-64313-3_25
Placements for a Humanoid Bipedal
Robot
Connor Morrow
(B) , Benjamin Bolen , and Alexander J. Hunt
Department of Mechanical and Materials Engineering, Portland State University,
Portland, OR 97207, USA
comorrow@pdx.edu
Abstract. This work demonstrates an algorithm that is able to compute
optimal placement for braided pneumatic actuators on a bipedal robot in
order to emulate the biology of human legs. The algorithm calculates the
torque that muscles are able to generate about a series of joints (back,
hip, knee, ankle, subtalar, and metatarsophalangeal) in a human model.
It then compares these torques to the torque that is achievable by a
reduced number of pneumatic muscles actuating a bipedal robot model
and optimizes the results to reduce the error between the robot and
human model. The algorithm successfully finds new muscle placements
that will be used in physical testing to verify that the torque output is
correct and matches similarly to human capabilities. The algorithm was
performed for three muscles about the back (lumbrosacral) joint: erector
spinae, internal oblique, and external oblique. It generates placements
capable of producing torque profiles that are more biologically realistic than the previously hand placed locations. Currently, the algorithm
is not prevented from placing muscle paths that intersect the physical
structure. This work will enable the development of controllable, physical models that more accurately capture force and torque capabilities of
human muscles. Such physical models will enable more complete testing
of how the nervous system performs effective control of over-actuated
muscle systems, and if such systems have advantages for robotic applications.
Keywords: Biomimetic robotics · Optimization algorithm ·
Pneumatic artificial muscle · Braided pneumatic actuator · Opensim ·
Matlab
1 Introduction
Biomimetic robots are an important topic of research to the robotics community,
due to their proposed ability to navigate through environments that are difficult for traditional robots. Robots have trouble navigating through spaces that
contain uneven surfaces or stairs, as well as areas in which there is a transition
c
Springer Nature Switzerland AG 2020
V. Vouloutsi et al. (Eds.): Living Machines 2020, LNAI 12413, pp. 257–269, 2020.
https://doi.org/10.1007/978-3-030-64313-3_25