Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.700817
Title: Morphological computation in active haptic embodied perception
Author: Sornkarn, Nantachai
ISNI:       0000 0004 5989 0212
Awarding Body: King's College London
Current Institution: King's College London (University of London)
Date of Award: 2016
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
This thesis presents a study on the role of internal impedance control in embodied perception. This study gives a novel perspective on how action and perception are coupled in a shared embodiment like a tendon in muscles used for both action and perception. The mode of perception discussed in this thesis is the sense of touch or “haptic perception” of both human and biologically inspired artificial systems. Firstly, this thesis explores the internal impedance control behavior of humans in haptic exploration during manual palpation of a soft phantom (presented by a soft silicone phantom) to detect and estimate the depth of an abnormality (presented by a hard nodule). The muscle actuation levels of humans were measured across human subjects to learn the pattern of such regulations. It was found that humans perform voluntary modulation of muscle co-contraction level during haptic exploration of a soft tissue. In addition, it was found that these regulations of muscle co-contraction can be learned and mapped using a Markov decision matrix. This raised the question, which became the main focus of this thesis, as to why humans perform such regulations of muscle co-contraction levels during haptic exploration. The influence of proprioceptive information and the muscular activity on the interpretation of the environment during haptic exploration is not understood yet. Therefore, the objective of this thesis is to understand the role of internal impedance and behavioral variables control during embodied sensing and haptic exploration. Secondly, it was found using a robotic manipulator with variable joint stiffness that the information-gain of the perception of its internal state can be maximized by controlling the joint’s stiffness (internal impedance). This leads to an enhanced accuracy in the estimation of its internal state. It was also found that the sensing of external environment, in this case, through haptic perception during robotic palpation could also benefit from this principle. The information gain about the environment can be maximized through the modulation of the internal stiffness. Thirdly, a Bayesian inference mechanism was used in addition to the information metrics in the robotic palpation task to infer real-time estimates of the depth of abnormality in soft silicone phantom based on past experience. The stiffness control pattern found in human’s manual palpation was implemented in the robotic probe to investigate whether controlling the probe’s internal impedance to follow the transition patterns of human’s co-contraction levels can enhance the nodule depth estimation accuracy. In comparison to the static stiffness; it was shown that this strategy of stiffness control in the robotic probe significantly improved the estimation accuracy of hard nodule’s depth. Lastly, this thesis also investigated both individual and collective roles of robotic probe’s internal stiffness, indentation level, and probe sweeping speed in the estimation of the nodule’s depth during haptic exploration. The results from the experiments have confirmed the hypothesis that by allowing the probe to vary its internal stiffness and behavioral variables, the estimation process can enhance the accuracy in haptic perception. Using artificial system as an abstraction of biological counterparts, this thesis has, for the first time, explained the possible reason as to why biological systems, humans, for instance, actuate both internal parameter (like stiffness of the joint) and the behavioral variables during haptic exploration of the environment.
Supervisor: Nanayakkara, Thrishantha ; Howard, Matthew Jacob William Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.700817  DOI: Not available
Share: