Robots often must achieve high speed to improve functionality and productivity.
Such speed may, however, cause higher internal or impact forces upon contact,
thus increasing the risk of damage to motors, gears, and other parts. Furthermore,
robots are composed of various mechanisms depending on their applications,
and their structural strengths will vary from moment to moment according
to the robot’s changing configurations or postures. Parts damaged by external
force will vary depending on construction, configuration, and movement.
There is always a possibility that some unexpected parts are damaged due
to stress concentration or the propagation of impact waves during robotic
motion. To prevent such damage, it is important to detect and mitigate
impact forces that propagate among structural components. In most cases,
sensors are used to detect impact forces, and thus the risk of structural
damage. Sensors, however, are limited in detecting impact only to parts
they measure. Accelerometers are often used in measuring shock waves, but
acceleration data cannot be used directly to evaluate structural risk since
the relationship between acceleration and impact force differs for different
materials. Although some methods employ force sensors, the measured data
is not reliable since calibration methods against dynamic loads remain
unestablished. Dynamic loads can only be measured precisely using momentum
change of the impacting body. Sensors and amplifiers have measurement limitations
against response periods, so robot may undergo high-frequency shock that
is undetected. Given these considerations, we propose a sensorfree robot
having a numerical model of its own mechanism and thus capable of predicting
impact forces upon contact with another body based on rapid finite element
analysis.
We investigated the possibility of constructing an impact force prediction
system for robotic mechanisms, by developing a scheme to analyze shock
waves occurring during robotic motion at low computational cost. We also
tested its validity by comparing the results with experiments. To stabilize
time integration and lower computational cost, we used linear Timoshenko
beam elements, which have a low number of dimensions and few degrees of
freedom (DOF), to model robots. Although linear Timoshenko beam elements
use linear interpolation as a displacement function and thus have low accuracy,
highly accurate elastic solutions are obtained by a minimum number of finite
elements by treating two elements as a single subset element and placing
its stress evaluation point at a location corresponding to the numerical
integration point of a Bernoulli-Euler beam element. Finite elements generally
place the mass at nodes, causing the mass at constrained points to be neglected.
We coped with this problem by concentrating the mass of subset elements
at nodes at centers of gravity. To treat contact with another object, we
used gap elements. Integrating these techniques, we constructed an internal
force distribution analysis scheme for a robotic mechanism, including its
contact with another object, and studied the validity of impact force analysis
by comparison with experimental results. We then applied the scheme to
analyze impact forces occurring during “bipedal motion” in a leg consisting
of stiffness varying members.
Outline of target motion
Experimental setup
Time histories of impact force
We developed a way of introducing gap elements to treat material properties and shape of the external contact object. We then determined the impact force generated in a 2-link mechanism. Peak impact forces agreed well with experimental results demonstrating that our proposal predicts potentially damaging contact to component materials.
Target motion of a biped robot
We constructed a simple model of a robotic leg mechanism for impact analysis. Using this model, we showed that MRF dampers, introduced as variable-stiffness components, effectively reduce impact forces. We also confirmed that the load carried by an individual leg is varied by changing member stiffness, and that inappropriate positioning of the MRF damper actually increases peak impact.
Related papers (Journals):
D.Isobe and Y.Moriya: A Finite Element Scheme for Impact Force Prediction of Robotic Mechanisms, Journal of Robotics and Mechatronics, Vol.18, No.3, (2006), pp.340-346. DOI: 10.20965/jrm.2006.p0340
Related papers (Proceedings):
K.Kozawa and D.Isobe: Shock Absorbing Analysis of Actively Transforming Architecture by Using Finite Element Method, CD-ROM Proceedings of the JSME Annual Conference on Robotics and Mechatronics '01, No.01-4, (2001), in Japanese. abstract
D.Isobe and K.Kozawa: Prediction of Impulsive Force in Walking Operation of Robotic Architecture by Using Finite Element Method, Proceedings of the 7th Robotics Symposia, (2002), pp.197-202, in Japanese. abstract
N.Hirota and D.Isobe: Prediction of Impact Force Produced in Robotic Mechanisms by Using FEM, CD-ROM Proceedings of the JSME Annual Conference on Robotics and Mechatronics '02, No.02-6, (2002), in Japanese. abstract
Y.Moriya, N.Hirota and D.Isobe: Impact analysis of a manipulator using FEM, CD-ROM Proceedings of the 21st Annual Conference of the Robotics Society of Japan 2003, (2003), in Japanese. abstract
D.Isobe, N.Hirota, Y.Moriya and N.Yamane: Impact-Force Prediction for Manipulators by Using Numerical Methods, CD-ROM Proceedings of the JSME Annual Conference on Robotics and Mechatronics '04, No.04-4, (2004), in Japanese. abstract
D.Isobe and N.Hirota: Impact-Force Prediction for Manipulators by Using Numerical Models, Proceedings of the Mechanical Engineering Congress Vol.1, No.04-1, (2004), pp.1-2, in Japanese. abstract
Y.Moriya and D.Isobe: Impact Force Identification of Link Mechanisms by Using Numerical Models, CD-ROM Proceedings of the 22nd Annual Conference of the Robotics Society of Japan 2004, (2004), in Japanese. abstract