Background

Nowadays most artificial (i.e., robotic or prosthetic) arms and hands are described as anthropomorphic without really evaluating their level of humanlikeness. The notion of Anthropomorphism comes from the Greek word anthropos that means human and the Greek word morphe that means form and according to Epley [1] its essence “is to imbue the imagined or real behavior of nonhuman agents with human-like characteristics, motivations, intentions and emotions”.

In the robotics literature, Mason et al. [2] discussed that the main reasons that researchers tend to be inspired by the human hand are the following:

• An anthropomorphic design interfaces well with human-centric environments (e.g., the objects and the environments surrounding us have been crafted for the human-hand).
• Human-likeness of robot hand designs provides intuitiveness for teleoperation studies.
• A humanlike design facilitates comparisons between biomechanical and robotic studies.
• A humanlike design is aesthetically advantageous for certain applications and environments (e.g., in prosthetic and entertainment oriented applications).
• For the simplest reason that the human hand, as an outcome of thousands of years of evolution is definitely a “good design”.

Motivation

Moreover, the last decade, anthropomorphism has received increased attention and has been a desired characteristic of robots for a plethora of Human - Robot Interaction applications for two main reasons:

• Robots need to establish solid social connections with humans and anthropomorphism is of paramount importance for this goal. After all it is well-known that the more humanlike a robot is in terms of appearance, motion and/or perceived intelligence, the more easily will manage to socially interact with humans, inducing them empathetic reactions.
• Anthropomorphism guarantees also safety of possible interactions between humans and robots (e.g., during a collaborative assembly task), as human-like motion can be more intuitively perceived by humans, which can comply their activities and/or motion, to avoid possible injuries.

The Toolbox

This repository contains a MATLAB toolbox for the quantification of structural anthropomorphism of artificial arms and hands. The toolbox is based on computational geometry and set theory methods and is distributed under an open-source license (GPLv3). For doing so, it compares human and artificial workspaces proposing a series of metrics that assess their relative coverages. For human and artificial arms the toolbox assesses anthropomorphism of the upper-arm, forearm and wrist / hand workspaces. For human and artificial hands the toolbox evaluates the anthropomorphism of the workspaces of the finger bases frames and the workspaces of the finger phalanges. Various methods for the computation of the workspaces volumes are also presented. The final score of anthropomorphism is derived as a weighted sum of the sub-scores of the independent workspaces metrics proposed and always ranges from 0 (for non-anthropomorphic artifacts) to 1 (for human-identical artifacts). Τhe toolbox can be used for design optimization of anthropomorphic prosthetic and robotic devices. More details for the quantification of anthropomorphism of artificial arms and hands, can be found in [3] and [4] respectively.

Fig. 1. Workspaces points and the corresponding convex hulls for the phalanges of the human index finger.

Fig. 2. Workspaces comparisons between human (black convex hulls) and two artificial arms (red convex hulls), for actual links lengths and normalized link lengths. Letter X denotes those cases for which no comparisons between the human and the artificial workspaces can be conducted (i.e., score of anthropomorphism is zero).

Fig. 3. Comparison of finger workspaces between a robot hand (black convex hulls) and the human hand (red convex hulls).

The toolbox is constantly under-development and the newest version can always be found in the OpenBionics/Anthropomorphism GitHub repository. We plan to eventually migrate the toolbox to Octave, R and/or Python (for people that do not have access to a MATLAB license). The toolbox has 3 main folders named: “Arm”, “Hand” and “Examples”. The first two folders contain the functions for the quantification of anthropomorphism of artificial arms and hands respectively, while the third folder contains examples of the implemented methods. Up to now (October 2015), the toolbox contains 55 functions and 3 examples. Should you have any questions or suggestions, feel free to contact us. Enjoy!

References

[1] N. Epley, A. Waytz, and J. T. Cacioppo, “On seeing human: a three-factor theory of anthropomorphism.,” Psychol. Rev., vol. 114, no. 4, pp. 864–886, 2007.

[2] M. T. Mason, A. Rodriguez, S. S. Srinivasa, and A. S. Vazquez, “Autonomous manipulation with a general-purpose simple hand,” The International Journal of Robotics Research, vol. 31. pp. 688–703, 2012.

[3] M. Liarokapis, P. Artemiadis, and K. Kyriakopoulos, “Quantifying Anthropomorphism of Robot Hands,” in IEEE International Conference on Robotics and Automation (ICRA), vol., no., pp.2041-2046, 6-10 May 2013.

[4] C. Mavrogiannis, M. V. Liarokapis, and K. J. Kyriakopoulos, “Quantifying Anthropomorphism of Robot Arms,” in IEEE/RSJ International Conference on Intelligent Robots and Systems, 2015.