Modulation of dynamic behavior of anthropomorphic robot: A biomimetic approach with force redundancy

Abstract

The biomimetic approach to robotics is an attempt to apply solutions created by the evolution of biological systems to technical systems. We investigate the feedforward modulation methodology of the dynamic behavior of an anthropomorphic manipulator, which inherently possess abundant actuators. In this work, we explain that for complete modulation of the dynamic behavior of such a system, redundant actuation is essential. Redundant actuation enables the system to modulate the stiffness or weighted stiffness properties such as motion frequency, and even modulates the damping ratio of the system. Thus, the modulation capability of the stiffness property plays an important role. We also address the need for the incorporation of active damper for complete modulation of the dynamic behavior. An illustrative example of a biomimetic, 2-degrees of freedom manipulator resembling the musculoskeletal structure of the human upper-extremity and driven by six actuators, is given. The modulation capability of motion frequencies and damping ratios is verified through a virtual trajectory planning that represents a point-to-point motion accomplished by the progressive movement of the equilibrium posture. To show the effectiveness of the proposed algorithms, several simulation results are illustrated.

Introduction

Biological systems solve problems that today’s technical systems cannot. Biomimetism implies an approach that mimics the intelligent structures and behaviors of biological systems. This approach has been applied to robotics [1], [2], [3], [4], [5], and it emphasizes the behavior of general autonomous systems. It has been proven that the study of animal behavior leads to models and hypotheses that help to create new technical solutions analogous to their successful counterparts in nature. The long-term goal is to create robots that autonomously interact for an unrestricted time with a real world environment and enhance their performance with the experience gained from ongoing interaction with the world.

Biologically inspired robotics often includes large numbers of actuators, heavily coupled skeletal structures and complex non-quantifiable performance parameters [6]. For example, a planar model of the human upper-extremity (arm) shown in Fig. 1 possesses six human actuators (i.e., muscles). These muscles seem to be redundant in a kinematic viewpoint and they are kinematically coupled by the complex skeletal structures. Developing a satisfactory control architecture for such anthropomorphic systems will be another challenging issue; although several algorithms, which were reported in redundantly actuated robotic systems, do exist [7], [8], [9], [10]. There will also be many design parameters in designing a similar robotic counterpart. One method of reducing the development time is by continuing to follow the biological analogy and develop a control architecture that can evolve to an acceptable level of performance.

Hogan [11] explained the existence of the hyper-redundant muscles of the human body in terms of the spring-like impedance property. Since the neural feedback control of dynamic behavior is curtailed by the inevitable time delays associated with the neural transmission, neural feedback is only effective below a certain frequency. Thus, Hogan noted that synergistic activation of antagonistically actuated redundant muscles might be a single unified approach to control dynamic interaction at all frequencies, and that the inertia property may be another important impedance property. Since the inertia property at the end-point of the human arm is a function of the joint angles of the system, the human arm with kinematically redundant structure is able to modulate the inertia property by its self-motion without changing the end-position of the arm. Inspired by those ideas, Yi and Freeman [12] developed a methodology for actively adjustable springs in general redundantly actuated closed-chain mechanisms. Also, this idea has been successfully applied to the analysis and design of redundantly actuated linkage systems [13], [14], [15], [16], [17] and some of the biomechanical models in terms of the spring-like impedance property [18, [19.

Studies on the mechanical characteristics and control model for the human arm have been carried out [20], [21], [22], [23]. Dolan et al. [20] estimated the dynamic impedance of the human arm by using an experimental methodology. They represented the stiffness–damping–mass characteristics numerically as matrices and graphically as ellipses that were characterized by size, shape, and orientation. They investigated the correlation between the dynamic components of the human arm and examined the ability to modulate stiffness in the face of initial loading. Koganezawa [21] investigated the task oriented stiffness of redundant manipulator similar to human upper-extremity. Lan [22] proposed a biological optimal control model for the generation of goal-directed multi-joint arm movements and studied the formation of muscle control inputs and invariant kinematic features of movements. The proposed model has a hierarchical structure that can determine the control inputs for a set of redundant muscles without any inverse computation. Stroeve [23] studied the mechanical impedance of neuromusculoskeletal models of the human arm. In his model, the motor control system is represented by a neural network that combines feedforward and feedback control. Thus, the achieved impedance characteristics depend on the conditions during the learning process. It can be observed that previous researchers have studied the arm model with six lumped muscles. The identical kinematic model given in Fig. 1 was employed in the previous research works [11], [20], [21], [22], [23]. However, they ignored the exact kinematic modeling of the model and the geometric analysis for the role of redundant actuation embedded in the human arm model.

Therefore, this work deals with beneficial application of the redundant actuation embedded in the musculoskeletal structure of the human arm model through the exact kinematic modeling and the geometric analysis for the human arm model. Based on this analysis, we explain that for the complete modulation of the dynamic behavior of anthropomorphic robotic systems, redundant actuation is essential to enable the system to modulate the stiffness or weighted stiffness properties such as motion frequency. We show that even in the modulation of the damping ratio of the system, the modulation capability of the stiffness property plays an important role. We also address the need for the incorporation of the active damper necessary for complete modulation of the dynamic behavior.

The organization of this paper is as follows: We explain the motivation of this work in Section 2. Kinematic and dynamic methodology for a given anthropomorphic robotic mechanism is included in Section 3. In Section 4, the motion frequency and damping ratio modulation algorithms and their associated load distribution scheme will be addressed. In Section 5, simulation results will be shown to demonstrate the effectiveness of the proposed algorithms. Finally, we draw conclusions.