137 research outputs found
Closed form solution to controller design for human-robot interaction
No full textThis paper deals with controller design for gentle physical
human-robot interaction. Two objectives are set up. The first is to
establish an analytical framework in order to justify the good
features of state of the art controller, recently designed by numerical search of parameter space. The second is to investigate the possibilities to improve the performance of such controller. Our method ensures “prescribed” admittance behavior of the robot, similar to natural admittance controller design but with both more realistic model of the robot and more realistic target admittance.
Joining natural admittance approach with the concept of complementary stability allows reaping the benefits of both. Limited knowledge about the environment via structured uncertainty allows a very simple worst-case analysis using elementary tools such as Routh–Hurwitz stability criterion. Consequent relation within the parameters determines an allowed region in the parameter space, where the contact stability is guaranteed. Not surprisingly, on one border of this region, the system behaves exactly the same as when the state of the art controller is employed. In addition, unexpected stability regions are discovered, suggesting theoretical performance improvements
Ectropy of diversity measures for populations in Euclidean space
No full textMeasures to evaluate the diversity of a set of points (population) in Euclidean space play an
important role in a variety of areas of science and engineering. Well-known measures are
often used without a clear insight into their quality and many of them do not appropriately
penalize populations with a few distant groups of collocated or closely located points. To
the best of our knowledge, there is a lack of rigorous criteria to compare diversity measures
and help select an appropriate one. In this work we define a mathematical notion of ectropy
for classifying diversity measures in terms of the extent to which they tend to penalize
point collocation, we investigate the advantages and disadvantages of several known measures
and we propose some novel ones that exhibit a good ectropic behavior. In particular,
we introduce a quasi-entropy measure based on a geometric covering problem, three measures
based on discrepancy from uniform distribution and one based on Euclidean minimum
spanning trees. All considered measures are tested and compared on a large set of
random and structured populations. Special attention is also devoted to the complexity
of computing the measures. Most of the novel measures compare favorably with the classical
ones in terms of ectropy. The measure based on Euclidean minimum spanning trees
turns out to be the most promising one in terms of the tradeoff between the ectropic
behavior and the computational complexity
Resilient hexapod robot
No full textIn this paper, we present a method of learning desired behaviour of the specific robotic system and transfer of the existing knowledge in the event of partial system failure. Six-legged robot (hexapod) built on top of the Bioloid platform is used for the method verification. We use genetic algorithms to optimize the hexapod's gait, after which we simulate physical damage caused to the robot. The goal of this method is to optimize the gait in accordance with the actual robot morphology, instead of the assumed one. Also, knowledge that was previously gained will be transferred in order to improve the results. Nonstandard genetic algorithm with the specific mixed population is used for thi
- …
