Supplementary material for:
Self-organized behavior generation for musculoskeletal robots

Ralf Der and Georg Martius

Under review. A preview version is on Arxiv: abs/1602.02990.

Abstract: With the accelerated development of robot technologies, optimal control becomes one of the central themes of research. In traditional approaches, the controller, by its internal functionality, finds appropriate actions on the basis of the history of sensor values, guided by the goals, intentions, objectives, learning schemes, and so on planted into it. The idea is that the controller controls the world—the body plus its environment—as reliably as possible. This paper advocates for a new paradigm of control, obtained by making the world control its controller in the first place. The paper presents a solution with a controller that is devoid of any functionalities of its own, given by a fixed, explicit and context-free function of the recent history of the sensor values. When applying this controller to a muscle-tendon driven arm-shoulder system from the Myorobotics toolkit, we observe a vast variety of self-organized behavior patterns: when left alone, the arm realizes pseudo-random sequences of different poses but one can also manipulate the system into definite motion patterns. But most interestingly, after attaching an object, the controller gets in a functional resonance with the object’s internal dynamics: when given a half-filled bottle, the system spontaneously starts shaking the bottle so that maximum response from the dynamics of the water is being generated. After attaching a pendulum to the arm, the controller drives the pendulum into a circular mode. In this way, the robot discovers affordances of objects its body is interacting with. We also discuss perspectives for using this controller paradigm for intention driven behavior generation.

1  Overview

The videos can be watched at

HandshakeHuman robot interaction by manually imposing a periodic movementVideo 1
Arm with pendulumSuspending a weight from the tip of the arm: self-excitation of a circular pendulum modeVideo 2
Pendulum responsesMotors are stopped. Recording spring forces of swinging suspended bottleVideo 3
Shaking horizontallyA half filled bottle is horizontally attached to the tip of the arm: shaking of the bottle mainly along its axisVideo 4
Shaking verticalVertical attachment, half filled: shaking direction mainly along the (now vertical) axisVideo 5
How to rotate a wheelArm attached frontally to a revolvable bar/wheel.Video 6
Rotating wheel IIParallel wheel – arm arrangementVideo 7
Wiping tableArm with brush starts to wipe a tableVideo 8
Wiping table modesDifferent wiping patterns from reloaded controllersVideo 9
Sensor disruptionsWith visual input for hand. Camera is turned during behavior. Fast reorganizationVideo 10
Hand-eye coordinationCoordination develops, such that arm follows a dummy handVideo 11

2  Videos

Video 1: Handshake: Human-robot interaction by manually imposing a periodic movement. A longer version can be found here. [mp4 video file]
back to List of Videos

Video 2: Bottle swing: Excitation of a circular pendulum mode. The suspended bottle, once exited to swing a little bit, excerts forces onto the arm, which are incoorporated into the controller throught the plasticity rule. This leads eventually to a coherent swinging motion. [mp4 video file]
back to List of Videos

Video 3: Pendulum responses: Motors are stopped. Recorded are the spring forces of a swinging suspended weight. See the figure in the paper for the resulting sensor readings. [mp4 video file]
back to List of Videos

Video 4: Shaking horizontally: A helf filled bottle is horizontally attached to the tip of the arm. The main shaking direction is horizontal. Stronge response to dynamics inside the bottle. [mp4 video file]
back to List of Videos

Video 5: Shaking vertically: A half filled bottle is vertically attached to the tip of the arm. Emergent shaking behavior of the bottle mainly along its axis. [mp4 video file]
back to List of Videos

Video 6: How to rotate a wheel: Arm is attached frontally to a revolvable bar/wheel. In the beginning, the arm is very loosely attached to the crank so that there is no definite force transfer. After improving the connection and some kick by the experimenter the arm rotates the wheel. It can then also quickly learn to rotate the other direction. [mp4 video file]
back to List of Videos

Video 7: Rotating wheel II: Parallel wheel – arm arrangement. The arm is self-learning to rotate the wheel and also quickly learns to rotate in the opposite direction. [mp4 video file]
back to List of Videos

Video 8: Wiping table: Arm with brush starts to wipe a table. The table plane and the manual guidance makes the to arm to wipe the table. Later in the video a different wiping pattern is generated. [mp4 video file]
back to List of Videos

Video 9: Wiping table modes: Different wiping patterns from reloaded controllers. Controller that where saved during a previous run where reloaded one after another. One observes smooth transitions between the different wiping modes, an example of the attractor morphing discussed in the paper. Occasionally, transitions may take some time. [mp4 video file]
back to List of Videos

Video 10: Including vision I: The camera coordinates for the green fist are used as additional sensor values. With fixed camera, cyclic arm motions emerge. When the camera is slowly rotated, these patterns are coherently morphing into a new pattern after the camera was being stopped. [mp4 video file]
back to List of Videos

Video 11: Including vision II: Emerging Fist-eye coordination: After a stable motion pattern has developed, the fist is capped but the camera now sees the dummy fist. Observe how the arm follows the dummy fist in a deliberate but irregular manner. However, the dummy can guide the arm in a coherent motion if it runs along the original trajectory. [mp4 video file]
back to List of Videos

This document was translated from LATEX by HEVEA.