{"id":{"repo_id":"mcmaster","oai_identifier":"oai:macsphere.mcmaster.ca:11375/33799"},"canonical_url":"https://search.ndltd.org/etd/mcmaster/oai:macsphere.mcmaster.ca:11375/33799","repository":{"repo_id":"mcmaster","name":"McMaster University","base_url":"https://macsphere.mcmaster.ca/server/oai/request"},"display":{"title":"Advancing The Perception-Action Cycle In Cognitive Dynamic Systems: Robust Estimation Methods","abstract":"This sandwich thesis comprises a comprehensive survey of cognitive dynamic systems (CDSs), followed by three technical contributions that advance the state of the art in robust state estimation for the perception stage of the perception-action cycle (PAC). Cognitive dynamic systems constitute an emerging class of engineering systems that draw inspiration from neuroscience, configuring traditional architectures around five cognitive processes (the perception-action cycle, memory, attention, intelligence, and language) to enable learning from experience and improved performance through continued interaction with the environment. The PAC sits at the foundation of this framework, since a CDS is only as capable as the perceptor that supplies state information to its downstream cognitive processes. In most implementations, the perceptor relies on the Kalman filter (KF) and its nonlinear extensions, which assume linear-Gaussian dynamics and an accurate system model. These assumptions are routinely violated in safety-critical applications by non-Gaussian noise, outliers, modelling uncertainty, and abrupt mode transitions induced by faults. While effective under nominal conditions, the KF and existing robust alternatives tend to address individual failure modes in isolation, and a unified framework that simultaneously confers robustness to non-Gaussian disturbances, tolerance of modelling uncertainty, and capability for fault detection in multi-mode systems, all without sacrificing nominal-case performance, has remained elusive. This thesis explores higher-order moments and variable structure correction as a unifying foundation for robust state estimation, augmented by entropy-regularized adaptation drawn from information theory. The proposed framework provides a mechanism for replacing the second-order statistical core of the KF with a correntropy-based criterion that captures all even-order moments of the estimation error, complemented by bounded sliding-mode-inspired corrective action and adaptive mechanisms that respond to the prevailing operating conditions. The overarching goal of this research is to establish a coherent family of estimators that strengthens the perceptor of a CDS while degrading gracefully to the conventional KF under nominal conditions. The research is organized around three technical objectives: (1) developing the Moments-based Kalman Filter (MKF) and Moments-based Innovation Filter (MIF), which unify correntropy-based and variable-structure-based robustness into a single estimation framework with adaptive kernel bandwidth driven by innovation skewness and kurtosis, supported by mathematical stability proofs and nonlinear extensions (EMKF, EMIF) for general dynamical systems, (2) introducing the Adaptive MIF (AMIF) by replacing the fixed boundary layer of the MIF with an entropy-regularized, time-varying formulation derived from an optimization that balances \\emph{a posteriori} estimation error against a log-determinant entropy penalty, yielding contraction under nominal conditions and expansion under disturbances without explicit fault detection or mode switching, and (3) integrating the moment-based filters into the interacting multiple model (IMM) framework, yielding the IMM-MIF estimator for simultaneous state estimation, mode classification, and fault diagnosis through moment-based likelihood evaluation. These contributions are validated on two electromechanical benchmark systems: a simulated electro-hydrostatic actuator (EHA) operating under leakage-induced faults, and an experimental magnetorheological (MR) damper exhibiting current-dependent multi-mode dynamics. Both systems were selected as representative of the nonlinear, fault-prone, multi-mode regime in which conventional estimators struggle. Computational and experimental results demonstrate that the proposed estimators consistently outperform the conventional KF, EKF, UKF, and their IMM variants, both in terms of state estimation accuracy and, in the multi-mode setting, mode classification accuracy. These results collectively underscore the potential of moment-based, adaptive, variable-structure estimation not only as a robust alternative to the Kalman filter but also as a versatile foundation for the perception stage of cognitive dynamic systems deployed in safety-critical applications. In the broader context of CDS research, these methods align with the overarching theme of the thesis to strengthen the perception-action cycle through estimators that are theoretically grounded in information theory, methodologically coherent across linear, nonlinear, adaptive, and multi-mode regimes, and applied to systems representative of real-world deployment conditions. By unifying correntropy and variable structure within a single adaptive framework, these approaches support the development of perceptors capable of operating reliably under non-Gaussian noise, modelling uncertainty, and abrupt fault-induced mode transitions, providing a more reliable foundation upon which the higher cognitive processes of memory, attention, intelligence, and language can operate. This thesis highlights the transformative potential of moment-based robust estimation for addressing the critical challenges associated with perception in cognitive dynamic systems. By leveraging higher-order statistical moments, variable structure correction, and entropy-regularized adaptation, the proposed estimators provide a robust, interpretable, and computationally tractable approach for state estimation in safety-critical applications. The findings lay a foundation for future research into adaptive moment-based estimation and its broader applications in cognitive dynamic systems and intelligent autonomous systems.","abstract_html":"This sandwich thesis comprises a comprehensive survey of cognitive dynamic systems (CDSs), followed by three technical contributions that advance the state of the art in robust state estimation for the perception stage of the perception-action cycle (PAC). Cognitive dynamic systems constitute an emerging class of engineering systems that draw inspiration from neuroscience, configuring traditional architectures around five cognitive processes (the perception-action cycle, memory, attention, intelligence, and language) to enable learning from experience and improved performance through continued interaction with the environment. The PAC sits at the foundation of this framework, since a CDS is only as capable as the perceptor that supplies state information to its downstream cognitive processes. In most implementations, the perceptor relies on the Kalman filter (KF) and its nonlinear extensions, which assume linear-Gaussian dynamics and an accurate system model. These assumptions are routinely violated in safety-critical applications by non-Gaussian noise, outliers, modelling uncertainty, and abrupt mode transitions induced by faults. While effective under nominal conditions, the KF and existing robust alternatives tend to address individual failure modes in isolation, and a unified framework that simultaneously confers robustness to non-Gaussian disturbances, tolerance of modelling uncertainty, and capability for fault detection in multi-mode systems, all without sacrificing nominal-case performance, has remained elusive. This thesis explores higher-order moments and variable structure correction as a unifying foundation for robust state estimation, augmented by entropy-regularized adaptation drawn from information theory. The proposed framework provides a mechanism for replacing the second-order statistical core of the KF with a correntropy-based criterion that captures all even-order moments of the estimation error, complemented by bounded sliding-mode-inspired corrective action and adaptive mechanisms that respond to the prevailing operating conditions. The overarching goal of this research is to establish a coherent family of estimators that strengthens the perceptor of a CDS while degrading gracefully to the conventional KF under nominal conditions. The research is organized around three technical objectives: (1) developing the Moments-based Kalman Filter (MKF) and Moments-based Innovation Filter (MIF), which unify correntropy-based and variable-structure-based robustness into a single estimation framework with adaptive kernel bandwidth driven by innovation skewness and kurtosis, supported by mathematical stability proofs and nonlinear extensions (EMKF, EMIF) for general dynamical systems, (2) introducing the Adaptive MIF (AMIF) by replacing the fixed boundary layer of the MIF with an entropy-regularized, time-varying formulation derived from an optimization that balances \\emph{a posteriori} estimation error against a log-determinant entropy penalty, yielding contraction under nominal conditions and expansion under disturbances without explicit fault detection or mode switching, and (3) integrating the moment-based filters into the interacting multiple model (IMM) framework, yielding the IMM-MIF estimator for simultaneous state estimation, mode classification, and fault diagnosis through moment-based likelihood evaluation. These contributions are validated on two electromechanical benchmark systems: a simulated electro-hydrostatic actuator (EHA) operating under leakage-induced faults, and an experimental magnetorheological (MR) damper exhibiting current-dependent multi-mode dynamics. Both systems were selected as representative of the nonlinear, fault-prone, multi-mode regime in which conventional estimators struggle. Computational and experimental results demonstrate that the proposed estimators consistently outperform the conventional KF, EKF, UKF, and their IMM variants, both in terms of state estimation accuracy and, in the multi-mode setting, mode classification accuracy. These results collectively underscore the potential of moment-based, adaptive, variable-structure estimation not only as a robust alternative to the Kalman filter but also as a versatile foundation for the perception stage of cognitive dynamic systems deployed in safety-critical applications. In the broader context of CDS research, these methods align with the overarching theme of the thesis to strengthen the perception-action cycle through estimators that are theoretically grounded in information theory, methodologically coherent across linear, nonlinear, adaptive, and multi-mode regimes, and applied to systems representative of real-world deployment conditions. By unifying correntropy and variable structure within a single adaptive framework, these approaches support the development of perceptors capable of operating reliably under non-Gaussian noise, modelling uncertainty, and abrupt fault-induced mode transitions, providing a more reliable foundation upon which the higher cognitive processes of memory, attention, intelligence, and language can operate. This thesis highlights the transformative potential of moment-based robust estimation for addressing the critical challenges associated with perception in cognitive dynamic systems. By leveraging higher-order statistical moments, variable structure correction, and entropy-regularized adaptation, the proposed estimators provide a robust, interpretable, and computationally tractable approach for state estimation in safety-critical applications. The findings lay a foundation for future research into adaptive moment-based estimation and its broader applications in cognitive dynamic systems and intelligent autonomous systems.","abstract_has_math":false,"creators":["Hilal, Waleed"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":"Mechanical Engineering","school":null,"contributors":[],"advisors":["Gadsden, S. Andrew"],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026","date_published":"2026","updated_at":"2026-11-18T19:37:36Z","subjects":[],"languages":["en"],"rights":["Attribution-NonCommercial-NoDerivatives 4.0 International"],"rights_urls":["http://creativecommons.org/licenses/by-nc-nd/4.0/"],"identifier_entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://doi.org/10.71548/25569"],"render_values":[{"text":"https://doi.org/10.71548/25569","href":"https://doi.org/10.71548/25569","code":true}]}],"person_identifier_entries":[]},"links":{"outbound_url":"https://hdl.handle.net/11375/33799","outbound_label":"Handle","outbound_source":"dc:identifier.uri","record_url":"https://hdl.handle.net/11375/33799","record_label":"Repository record","record_kind":"handle","record_source":"dc:identifier.uri","download_url":null,"download_label":null,"download_source":null,"access_status":null,"embargo_until":null,"access_note":null},"source_record":{"url":"https://macsphere.mcmaster.ca/server/oai/request?verb=GetRecord&metadataPrefix=dim&identifier=oai%3Amacsphere.mcmaster.ca%3A11375%2F33799","prefix":"dim"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Gadsden, S. Andrew"]},{"key":"dc:contributor.department","label":"Department","values":["Mechanical Engineering"]},{"key":"dc:creator","label":"Author","values":["Hilal, Waleed"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2026-10-14T14:30:00Z"]},{"key":"dc:date.issued","label":"Date","values":["2026"]},{"key":"dc:type","label":"Dc Type","values":["Dissertation","Thesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]},{"key":"dc:rights","label":"Dc Rights","values":["Attribution-NonCommercial-NoDerivatives 4.0 International"]},{"key":"dc:rights.uri","label":"Rights URI","values":["http://creativecommons.org/licenses/by-nc-nd/4.0/"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["https://hdl.handle.net/11375/33799","https://doi.org/10.71548/25569"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["This sandwich thesis comprises a comprehensive survey of cognitive dynamic systems (CDSs), followed by three technical contributions that advance the state of the art in robust state estimation for the perception stage of the perception-action cycle (PAC). Cognitive dynamic systems constitute an emerging class of engineering systems that draw inspiration from neuroscience, configuring traditional architectures around five cognitive processes (the perception-action cycle, memory, attention, intelligence, and language) to enable learning from experience and improved performance through continued interaction with the environment. The PAC sits at the foundation of this framework, since a CDS is only as capable as the perceptor that supplies state information to its downstream cognitive processes. In most implementations, the perceptor relies on the Kalman filter (KF) and its nonlinear extensions, which assume linear-Gaussian dynamics and an accurate system model. These assumptions are routinely violated in safety-critical applications by non-Gaussian noise, outliers, modelling uncertainty, and abrupt mode transitions induced by faults. While effective under nominal conditions, the KF and existing robust alternatives tend to address individual failure modes in isolation, and a unified framework that simultaneously confers robustness to non-Gaussian disturbances, tolerance of modelling uncertainty, and capability for fault detection in multi-mode systems, all without sacrificing nominal-case performance, has remained elusive. This thesis explores higher-order moments and variable structure correction as a unifying foundation for robust state estimation, augmented by entropy-regularized adaptation drawn from information theory. The proposed framework provides a mechanism for replacing the second-order statistical core of the KF with a correntropy-based criterion that captures all even-order moments of the estimation error, complemented by bounded sliding-mode-inspired corrective action and adaptive mechanisms that respond to the prevailing operating conditions. The overarching goal of this research is to establish a coherent family of estimators that strengthens the perceptor of a CDS while degrading gracefully to the conventional KF under nominal conditions. The research is organized around three technical objectives: (1) developing the Moments-based Kalman Filter (MKF) and Moments-based Innovation Filter (MIF), which unify correntropy-based and variable-structure-based robustness into a single estimation framework with adaptive kernel bandwidth driven by innovation skewness and kurtosis, supported by mathematical stability proofs and nonlinear extensions (EMKF, EMIF) for general dynamical systems, (2) introducing the Adaptive MIF (AMIF) by replacing the fixed boundary layer of the MIF with an entropy-regularized, time-varying formulation derived from an optimization that balances \\emph{a posteriori} estimation error against a log-determinant entropy penalty, yielding contraction under nominal conditions and expansion under disturbances without explicit fault detection or mode switching, and (3) integrating the moment-based filters into the interacting multiple model (IMM) framework, yielding the IMM-MIF estimator for simultaneous state estimation, mode classification, and fault diagnosis through moment-based likelihood evaluation. These contributions are validated on two electromechanical benchmark systems: a simulated electro-hydrostatic actuator (EHA) operating under leakage-induced faults, and an experimental magnetorheological (MR) damper exhibiting current-dependent multi-mode dynamics. Both systems were selected as representative of the nonlinear, fault-prone, multi-mode regime in which conventional estimators struggle. Computational and experimental results demonstrate that the proposed estimators consistently outperform the conventional KF, EKF, UKF, and their IMM variants, both in terms of state estimation accuracy and, in the multi-mode setting, mode classification accuracy. These results collectively underscore the potential of moment-based, adaptive, variable-structure estimation not only as a robust alternative to the Kalman filter but also as a versatile foundation for the perception stage of cognitive dynamic systems deployed in safety-critical applications. In the broader context of CDS research, these methods align with the overarching theme of the thesis to strengthen the perception-action cycle through estimators that are theoretically grounded in information theory, methodologically coherent across linear, nonlinear, adaptive, and multi-mode regimes, and applied to systems representative of real-world deployment conditions. By unifying correntropy and variable structure within a single adaptive framework, these approaches support the development of perceptors capable of operating reliably under non-Gaussian noise, modelling uncertainty, and abrupt fault-induced mode transitions, providing a more reliable foundation upon which the higher cognitive processes of memory, attention, intelligence, and language can operate. This thesis highlights the transformative potential of moment-based robust estimation for addressing the critical challenges associated with perception in cognitive dynamic systems. By leveraging higher-order statistical moments, variable structure correction, and entropy-regularized adaptation, the proposed estimators provide a robust, interpretable, and computationally tractable approach for state estimation in safety-critical applications. The findings lay a foundation for future research into adaptive moment-based estimation and its broader applications in cognitive dynamic systems and intelligent autonomous systems."]},{"key":"dc:description.degree","label":"Dc Description Degree","values":["Doctor of Philosophy (PhD)"]},{"key":"dc:title","label":"Title","values":["Advancing The Perception-Action Cycle In Cognitive Dynamic Systems: Robust Estimation Methods"]},{"key":"others:access_status","label":"Others Access Status","values":["embargo"]}]}],"canonical_facts":{"dc:contributor.advisor":["Gadsden, S. Andrew"],"dc:contributor.department":["Mechanical Engineering"],"dc:creator":["Hilal, Waleed"],"dc:date.accessioned":["2026-10-14T14:30:00Z"],"dc:date.issued":["2026"],"dc:description.abstract":["This sandwich thesis comprises a comprehensive survey of cognitive dynamic systems (CDSs), followed by three technical contributions that advance the state of the art in robust state estimation for the perception stage of the perception-action cycle (PAC). Cognitive dynamic systems constitute an emerging class of engineering systems that draw inspiration from neuroscience, configuring traditional architectures around five cognitive processes (the perception-action cycle, memory, attention, intelligence, and language) to enable learning from experience and improved performance through continued interaction with the environment. The PAC sits at the foundation of this framework, since a CDS is only as capable as the perceptor that supplies state information to its downstream cognitive processes. In most implementations, the perceptor relies on the Kalman filter (KF) and its nonlinear extensions, which assume linear-Gaussian dynamics and an accurate system model. These assumptions are routinely violated in safety-critical applications by non-Gaussian noise, outliers, modelling uncertainty, and abrupt mode transitions induced by faults. While effective under nominal conditions, the KF and existing robust alternatives tend to address individual failure modes in isolation, and a unified framework that simultaneously confers robustness to non-Gaussian disturbances, tolerance of modelling uncertainty, and capability for fault detection in multi-mode systems, all without sacrificing nominal-case performance, has remained elusive. This thesis explores higher-order moments and variable structure correction as a unifying foundation for robust state estimation, augmented by entropy-regularized adaptation drawn from information theory. The proposed framework provides a mechanism for replacing the second-order statistical core of the KF with a correntropy-based criterion that captures all even-order moments of the estimation error, complemented by bounded sliding-mode-inspired corrective action and adaptive mechanisms that respond to the prevailing operating conditions. The overarching goal of this research is to establish a coherent family of estimators that strengthens the perceptor of a CDS while degrading gracefully to the conventional KF under nominal conditions. The research is organized around three technical objectives: (1) developing the Moments-based Kalman Filter (MKF) and Moments-based Innovation Filter (MIF), which unify correntropy-based and variable-structure-based robustness into a single estimation framework with adaptive kernel bandwidth driven by innovation skewness and kurtosis, supported by mathematical stability proofs and nonlinear extensions (EMKF, EMIF) for general dynamical systems, (2) introducing the Adaptive MIF (AMIF) by replacing the fixed boundary layer of the MIF with an entropy-regularized, time-varying formulation derived from an optimization that balances \\emph{a posteriori} estimation error against a log-determinant entropy penalty, yielding contraction under nominal conditions and expansion under disturbances without explicit fault detection or mode switching, and (3) integrating the moment-based filters into the interacting multiple model (IMM) framework, yielding the IMM-MIF estimator for simultaneous state estimation, mode classification, and fault diagnosis through moment-based likelihood evaluation. These contributions are validated on two electromechanical benchmark systems: a simulated electro-hydrostatic actuator (EHA) operating under leakage-induced faults, and an experimental magnetorheological (MR) damper exhibiting current-dependent multi-mode dynamics. Both systems were selected as representative of the nonlinear, fault-prone, multi-mode regime in which conventional estimators struggle. Computational and experimental results demonstrate that the proposed estimators consistently outperform the conventional KF, EKF, UKF, and their IMM variants, both in terms of state estimation accuracy and, in the multi-mode setting, mode classification accuracy. These results collectively underscore the potential of moment-based, adaptive, variable-structure estimation not only as a robust alternative to the Kalman filter but also as a versatile foundation for the perception stage of cognitive dynamic systems deployed in safety-critical applications. In the broader context of CDS research, these methods align with the overarching theme of the thesis to strengthen the perception-action cycle through estimators that are theoretically grounded in information theory, methodologically coherent across linear, nonlinear, adaptive, and multi-mode regimes, and applied to systems representative of real-world deployment conditions. By unifying correntropy and variable structure within a single adaptive framework, these approaches support the development of perceptors capable of operating reliably under non-Gaussian noise, modelling uncertainty, and abrupt fault-induced mode transitions, providing a more reliable foundation upon which the higher cognitive processes of memory, attention, intelligence, and language can operate. This thesis highlights the transformative potential of moment-based robust estimation for addressing the critical challenges associated with perception in cognitive dynamic systems. By leveraging higher-order statistical moments, variable structure correction, and entropy-regularized adaptation, the proposed estimators provide a robust, interpretable, and computationally tractable approach for state estimation in safety-critical applications. The findings lay a foundation for future research into adaptive moment-based estimation and its broader applications in cognitive dynamic systems and intelligent autonomous systems."],"dc:description.degree":["Doctor of Philosophy (PhD)"],"dc:identifier.uri":["https://hdl.handle.net/11375/33799","https://doi.org/10.71548/25569"],"dc:language.iso":["en"],"dc:rights":["Attribution-NonCommercial-NoDerivatives 4.0 International"],"dc:rights.uri":["http://creativecommons.org/licenses/by-nc-nd/4.0/"],"dc:title":["Advancing The Perception-Action Cycle In Cognitive Dynamic Systems: Robust Estimation Methods"],"dc:type":["Dissertation","Thesis"],"others:access_status":["embargo"]},"updated_at":"2026-11-18T19:37:36Z"}