Iterative Learning Control
as a Framework for Human-Inspired
Control with Bio-mimetic Actuators
Franco Angelini
1,2(B) , Matteo Bianchi
1 , Manolo Garabini
1 ,
Antonio Bicchi
1,2 , and Cosimo Della Santina
3,4,5
1 Centro di Ricerca “Enrico Piaggio” and DII, Universit` a di Pisa, Pisa, Italy
frncangelini@gmail.com
2 Soft Robotics for Human Cooperation and Rehabilitation, IIT, Genova, Italy
3 Institute of Robotics and Mechatronics, DLR, Oberpfaffenhofen, Weßling, Germany
4 Department of Informatics, Technical University Munich, Garching, Germany
5 Cognitive Robotics Department,
Delft University of Technology, Delft, The Netherlands
Abstract. The synergy between musculoskeletal and central nervous
systems empowers humans to achieve a high level of motor performance, which is still unmatched in bio-inspired robotic systems. Literature already presents a wide range of robots that mimic the human
body. However, under a control point of view, substantial advancements
are still needed to fully exploit the new possibilities provided by these
systems. In this paper, we test experimentally that an Iterative Learning Control algorithm can be used to reproduce functionalities of the
human central nervous system - i.e. learning by repetition, after-effect
on known trajectories and anticipatory behavior - while controlling a
bio-mimetically actuated robotic arm.
Keywords: Motion and motor control · Natural machine motion ·
Human-inspired control
1 Introduction
Natural and bio-inspired robot bodies are complex systems, characterized by an
unknown nonlinear dynamics and redundancy of degrees of freedom (DoFs). This
poses considerable challenges for standard control techniques. For this reason,
researchers started taking inspiration from the effective Central Nervous System
(CNS), when designing controllers for robots [4,5]. In this work, we test experimentally a model-free controller intended for trajectory tracking with biomimetic
robots. We prove that the required tracking performances can be matched, while
presenting well-known characteristics of human motor control system, i.e. learning by repetition, mirror-image aftereffect, and anticipatory behavior. We do
that by presenting experiments on a robotic arm with two degrees of freedom,
each of which is actuated by means of a bio-mimetic mechanism replicating the
behavior of a pair of human muscles [7] (Fig. 1(a)).
c
Springer Nature Switzerland AG 2020
V. Vouloutsi et al. (Eds.): Living Machines 2020, LNAI 12413, pp. 12–16, 2020.
https://doi.org/10.1007/978-3-030-64313-3_2
as a Framework for Human-Inspired
Control with Bio-mimetic Actuators
Franco Angelini
1,2(B) , Matteo Bianchi
1 , Manolo Garabini
1 ,
Antonio Bicchi
1,2 , and Cosimo Della Santina
3,4,5
1 Centro di Ricerca “Enrico Piaggio” and DII, Universit` a di Pisa, Pisa, Italy
frncangelini@gmail.com
2 Soft Robotics for Human Cooperation and Rehabilitation, IIT, Genova, Italy
3 Institute of Robotics and Mechatronics, DLR, Oberpfaffenhofen, Weßling, Germany
4 Department of Informatics, Technical University Munich, Garching, Germany
5 Cognitive Robotics Department,
Delft University of Technology, Delft, The Netherlands
Abstract. The synergy between musculoskeletal and central nervous
systems empowers humans to achieve a high level of motor performance, which is still unmatched in bio-inspired robotic systems. Literature already presents a wide range of robots that mimic the human
body. However, under a control point of view, substantial advancements
are still needed to fully exploit the new possibilities provided by these
systems. In this paper, we test experimentally that an Iterative Learning Control algorithm can be used to reproduce functionalities of the
human central nervous system - i.e. learning by repetition, after-effect
on known trajectories and anticipatory behavior - while controlling a
bio-mimetically actuated robotic arm.
Keywords: Motion and motor control · Natural machine motion ·
Human-inspired control
1 Introduction
Natural and bio-inspired robot bodies are complex systems, characterized by an
unknown nonlinear dynamics and redundancy of degrees of freedom (DoFs). This
poses considerable challenges for standard control techniques. For this reason,
researchers started taking inspiration from the effective Central Nervous System
(CNS), when designing controllers for robots [4,5]. In this work, we test experimentally a model-free controller intended for trajectory tracking with biomimetic
robots. We prove that the required tracking performances can be matched, while
presenting well-known characteristics of human motor control system, i.e. learning by repetition, mirror-image aftereffect, and anticipatory behavior. We do
that by presenting experiments on a robotic arm with two degrees of freedom,
each of which is actuated by means of a bio-mimetic mechanism replicating the
behavior of a pair of human muscles [7] (Fig. 1(a)).
c
Springer Nature Switzerland AG 2020
V. Vouloutsi et al. (Eds.): Living Machines 2020, LNAI 12413, pp. 12–16, 2020.
https://doi.org/10.1007/978-3-030-64313-3_2