Pub Date : 2016-07-06DOI: 10.1109/ACC.2016.7525106
Yucong Lin, S. Saripalli
We present the design and implementation of a path planning algorithm for Unmanned Aerial Vehicles (UAVs) to avoid collisions with other aircraft. The aircraft are indifferent to the UAV. A sampling based method is developed to generate the avoidance path. The UAV's closed-loop system is used to simulate the trajectory in collision check used in path planning. The generated path is then checked against updated states of the UAV and obstacle aircraft, and a new path will be planned if the original one leads to a collision. The algorithm is validated in Hardware-In-the-Loop simulation (HIL) and real flight experiments. The paper shows that the algorithm is able to generate paths for the UAV to avoid obstacles of different numbers, approaching angles, and speeds.
{"title":"Sampling based collision avoidance for UAVs","authors":"Yucong Lin, S. Saripalli","doi":"10.1109/ACC.2016.7525106","DOIUrl":"https://doi.org/10.1109/ACC.2016.7525106","url":null,"abstract":"We present the design and implementation of a path planning algorithm for Unmanned Aerial Vehicles (UAVs) to avoid collisions with other aircraft. The aircraft are indifferent to the UAV. A sampling based method is developed to generate the avoidance path. The UAV's closed-loop system is used to simulate the trajectory in collision check used in path planning. The generated path is then checked against updated states of the UAV and obstacle aircraft, and a new path will be planned if the original one leads to a collision. The algorithm is validated in Hardware-In-the-Loop simulation (HIL) and real flight experiments. The paper shows that the algorithm is able to generate paths for the UAV to avoid obstacles of different numbers, approaching angles, and speeds.","PeriodicalId":137983,"journal":{"name":"2016 American Control Conference (ACC)","volume":"87 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123174130","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-07-06DOI: 10.1109/ACC.2016.7525505
Zhengyuan Zhou, B. Yolken, R. A. Miura-Ko, N. Bambos
An influence network consists of a set of interacting agents, each of whose actions produces effects or influence on his neighbors' actions. In general, the effects can be arbitrary functions of the neighbors' joint action; and they can be inhomogeneous among agents. Such effects are akin to the externalities (positive or negative) engendered on the action level (which further impacts one's utilities), and capture the commonality of a variety of networks in economics and engineering. Therefore, the study of strategic interactions among agents in an influence network and a characterization of an equilibrium can be of great value and applicability. In this paper, we formulate a simple game-theoretical model of influence networks that aims to study strategic interactions among agents in light of such influence. We then establish a connection between the resulting multi-player game with the well-known nonlinear complementarity problem (NCP). This connection not only places our work in the existing literature of a well-studied subject, thus deepening the understanding of both problems, but also, to a certain extent, allows us to leverage the tools in the NCP literature to identify different sufficient conditions for the existence and uniqueness of a Nash equilibrium in the multi-player game. We then characterize two broadly-scoped classes of influence networks with natural and intuitive interpretations and draw tools from fixed point theory to show the existence of a Nash equilibrium, as the existing results in the NCP literature are not directly applicable. We finally comment on the distributed, model-agnostic best response dynamics and show that under certain cases, they converge to a Nash equilibrium.
{"title":"A game-theoretical formulation of influence networks","authors":"Zhengyuan Zhou, B. Yolken, R. A. Miura-Ko, N. Bambos","doi":"10.1109/ACC.2016.7525505","DOIUrl":"https://doi.org/10.1109/ACC.2016.7525505","url":null,"abstract":"An influence network consists of a set of interacting agents, each of whose actions produces effects or influence on his neighbors' actions. In general, the effects can be arbitrary functions of the neighbors' joint action; and they can be inhomogeneous among agents. Such effects are akin to the externalities (positive or negative) engendered on the action level (which further impacts one's utilities), and capture the commonality of a variety of networks in economics and engineering. Therefore, the study of strategic interactions among agents in an influence network and a characterization of an equilibrium can be of great value and applicability. In this paper, we formulate a simple game-theoretical model of influence networks that aims to study strategic interactions among agents in light of such influence. We then establish a connection between the resulting multi-player game with the well-known nonlinear complementarity problem (NCP). This connection not only places our work in the existing literature of a well-studied subject, thus deepening the understanding of both problems, but also, to a certain extent, allows us to leverage the tools in the NCP literature to identify different sufficient conditions for the existence and uniqueness of a Nash equilibrium in the multi-player game. We then characterize two broadly-scoped classes of influence networks with natural and intuitive interpretations and draw tools from fixed point theory to show the existence of a Nash equilibrium, as the existing results in the NCP literature are not directly applicable. We finally comment on the distributed, model-agnostic best response dynamics and show that under certain cases, they converge to a Nash equilibrium.","PeriodicalId":137983,"journal":{"name":"2016 American Control Conference (ACC)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121814457","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-07-06DOI: 10.1109/ACC.2016.7525461
C. Somarakis, Dipankar Maity, J. Baras
We consider the Economic Dispatch Problem (EDP) in power systems for a smart-grid friendly environment. We develop a consensus based decentralized optimization algorithm that evolves in a time-varying communication network that suffers from multiple propagation time-dependent delays. This paper, being an improvisation of an earlier work of ours, addresses the effects of time dependent delays in a fully decentralized network. We show analytically and by simulation that propagation delays may not only affect the performance of the dynamic algorithm that solves the EDP but can also destabilize the system.
{"title":"The effect of delays in the Economic Dispatch Problem for smart grid architectures","authors":"C. Somarakis, Dipankar Maity, J. Baras","doi":"10.1109/ACC.2016.7525461","DOIUrl":"https://doi.org/10.1109/ACC.2016.7525461","url":null,"abstract":"We consider the Economic Dispatch Problem (EDP) in power systems for a smart-grid friendly environment. We develop a consensus based decentralized optimization algorithm that evolves in a time-varying communication network that suffers from multiple propagation time-dependent delays. This paper, being an improvisation of an earlier work of ours, addresses the effects of time dependent delays in a fully decentralized network. We show analytically and by simulation that propagation delays may not only affect the performance of the dynamic algorithm that solves the EDP but can also destabilize the system.","PeriodicalId":137983,"journal":{"name":"2016 American Control Conference (ACC)","volume":"90 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116652005","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-07-06DOI: 10.1109/ACC.2016.7525617
Y. Kobayashi, Kazuaki Sakurai, N. Yamada
In order to estimate the critical temperature ratio (CTR) of thermoacoustic engines, a steady-state oscillation control method is proposed. The amplitude of the oscillating pressure is regulated to a reference value by a proportional-integral (PI) controller and a low-pass filter without specifying oscillation frequency. The output signal of the PI controller is used as a time-varying gain in a phase-delay controller so that the plant energy is maintained at a constant that depends on the reference pressure amplitude. A linear relation between the steady-state gain and temperature ratio is empirically observed, which provides an alternative method for estimating the CTR. The closed-loop stability of the proposed control system is analyzed by assuming a second-order resonance model, from which a stability condition on the PI controller's gains and the cut-off frequency of the low-pass filter is derived. The validity of the derived stability condition is shown by experimental results and numerical simulations.
{"title":"Constant energy control by time-varying gain for steady-state oscillation of thermoacoustic engines to estimate critical temperature ratio","authors":"Y. Kobayashi, Kazuaki Sakurai, N. Yamada","doi":"10.1109/ACC.2016.7525617","DOIUrl":"https://doi.org/10.1109/ACC.2016.7525617","url":null,"abstract":"In order to estimate the critical temperature ratio (CTR) of thermoacoustic engines, a steady-state oscillation control method is proposed. The amplitude of the oscillating pressure is regulated to a reference value by a proportional-integral (PI) controller and a low-pass filter without specifying oscillation frequency. The output signal of the PI controller is used as a time-varying gain in a phase-delay controller so that the plant energy is maintained at a constant that depends on the reference pressure amplitude. A linear relation between the steady-state gain and temperature ratio is empirically observed, which provides an alternative method for estimating the CTR. The closed-loop stability of the proposed control system is analyzed by assuming a second-order resonance model, from which a stability condition on the PI controller's gains and the cut-off frequency of the low-pass filter is derived. The validity of the derived stability condition is shown by experimental results and numerical simulations.","PeriodicalId":137983,"journal":{"name":"2016 American Control Conference (ACC)","volume":"107 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117247480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-07-06DOI: 10.1109/ACC.2016.7525153
M. Demetriou
This paper considers the state estimation of spatially distributed processes via the employment of mobile sensors. In order to reduce the computational demands, a domain decomposition method is utilized and which decomposes the spatial domain into two subdomains. The estimator in the inner domain implements a hybrid Kalman filter with a mobile sensor, whereas the outer subdomain implements a naive observer. Coupling conditions at the inner/outer boundary serve as a means to exchange information between the two estimators and which constitute a consensus protocol. The motion of the mobile sensor is based on a spatial gradient scheme and serves as a further means of reducing the computational load associated with the solution to large scale differential Riccati equations. The proposed scheme is further extended to collaborative distributed estimation in which multiple mobile sensors are enforcing another level of consensus protocol in order to penalize the differences between their state estimates. Extensive numerical simulations of a one-dimensional parabolic partial differential equation are presented to further demonstrate the multi-level computational savings associated with the use of domain decomposition in state estimation of spatially distributed processes with mobile sensors.
{"title":"Domain decomposition methods in the distributed estimation of spatially distributed processes with mobile sensors","authors":"M. Demetriou","doi":"10.1109/ACC.2016.7525153","DOIUrl":"https://doi.org/10.1109/ACC.2016.7525153","url":null,"abstract":"This paper considers the state estimation of spatially distributed processes via the employment of mobile sensors. In order to reduce the computational demands, a domain decomposition method is utilized and which decomposes the spatial domain into two subdomains. The estimator in the inner domain implements a hybrid Kalman filter with a mobile sensor, whereas the outer subdomain implements a naive observer. Coupling conditions at the inner/outer boundary serve as a means to exchange information between the two estimators and which constitute a consensus protocol. The motion of the mobile sensor is based on a spatial gradient scheme and serves as a further means of reducing the computational load associated with the solution to large scale differential Riccati equations. The proposed scheme is further extended to collaborative distributed estimation in which multiple mobile sensors are enforcing another level of consensus protocol in order to penalize the differences between their state estimates. Extensive numerical simulations of a one-dimensional parabolic partial differential equation are presented to further demonstrate the multi-level computational savings associated with the use of domain decomposition in state estimation of spatially distributed processes with mobile sensors.","PeriodicalId":137983,"journal":{"name":"2016 American Control Conference (ACC)","volume":"82 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121118560","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-07-06DOI: 10.1109/ACC.2016.7524939
H. Evain, M. Rognant, D. Alazard, J. Mignot
This paper presents an allocator for over-actuated systems based on the Extended Kalman Filter (EKF). The main advantages of the proposed approach are the greater flexibility in handling the constraints and its real-time capabilities. Based on the literature, theoretical convergence results, which ensure the convergence towards the local optimal values looked for, are presented. Another formulation of the kinematic equations of redundant systems that meet some constraints is also proposed in order to go through and/or avoid singularities. The two formulations are combined and applied to an academic example (a planar redundant manipulator arm).
{"title":"Nonlinear dynamic inversion for redundant systems using the EKF formalism","authors":"H. Evain, M. Rognant, D. Alazard, J. Mignot","doi":"10.1109/ACC.2016.7524939","DOIUrl":"https://doi.org/10.1109/ACC.2016.7524939","url":null,"abstract":"This paper presents an allocator for over-actuated systems based on the Extended Kalman Filter (EKF). The main advantages of the proposed approach are the greater flexibility in handling the constraints and its real-time capabilities. Based on the literature, theoretical convergence results, which ensure the convergence towards the local optimal values looked for, are presented. Another formulation of the kinematic equations of redundant systems that meet some constraints is also proposed in order to go through and/or avoid singularities. The two formulations are combined and applied to an academic example (a planar redundant manipulator arm).","PeriodicalId":137983,"journal":{"name":"2016 American Control Conference (ACC)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121127637","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-07-06DOI: 10.1109/ACC.2016.7525517
V. Preciado, M. Amin Rahimian
We propose a theoretical framework to study the eigenvalue spectra of the controllability Gramian of systems with random state matrices, such as networked systems with a random graph structure. Using random matrix theory, we provide expressions for the moments of the eigenvalue distribution of the controllability Gramian. These moments can then be used to derive useful properties of the eigenvalue distribution of the Gramian (in some cases, even closed-form expressions for the distribution). We illustrate this framework by considering system matrices derived from common random graph and matrix ensembles, such as the Wigner ensemble, the Gaussian Orthogonal Ensemble (GOE), and random regular graphs. Subsequently, we illustrate how the eigenvalue distribution of the Gramian can be used to draw conclusions about the energy required to control random system.
{"title":"Controllability Gramian spectra of random networks","authors":"V. Preciado, M. Amin Rahimian","doi":"10.1109/ACC.2016.7525517","DOIUrl":"https://doi.org/10.1109/ACC.2016.7525517","url":null,"abstract":"We propose a theoretical framework to study the eigenvalue spectra of the controllability Gramian of systems with random state matrices, such as networked systems with a random graph structure. Using random matrix theory, we provide expressions for the moments of the eigenvalue distribution of the controllability Gramian. These moments can then be used to derive useful properties of the eigenvalue distribution of the Gramian (in some cases, even closed-form expressions for the distribution). We illustrate this framework by considering system matrices derived from common random graph and matrix ensembles, such as the Wigner ensemble, the Gaussian Orthogonal Ensemble (GOE), and random regular graphs. Subsequently, we illustrate how the eigenvalue distribution of the Gramian can be used to draw conclusions about the energy required to control random system.","PeriodicalId":137983,"journal":{"name":"2016 American Control Conference (ACC)","volume":"12 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127134737","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-07-06DOI: 10.1109/ACC.2016.7525284
Xin Xin, Takanori Kikkawa, Yannian Liu
This paper concerns the problem of how to obtain analytically all equilibrium points of the standard Kuramoto model (with all-to-all uniform coupling) having any natural frequencies and coupling gain. As an initial effort to solve this challenging problem, for the Kuramoto model with 3 (respectively 4) oscillators, to obtain analytically all equilibrium points, we show that we need to solve a polynomial equation of the sine of the phase of an oscillator with the highest order of 6 (respectively 14). For 3 oscillators, this polynomial equation with numerical examples shows that the maximal number of distinct equilibrium points for any natural frequencies and coupling gain is 6. For 4 oscillators, this paper shows theoretically that the maximal number of distinct equilibrium points is not greater than 14, and presents two numerical examples to show the existence of 10 distinct equilibrium points. From the numerical investigation carried out in this study, it is a conjecture that the maximal number of distinct equilibrium points of 4 oscillators for all natural frequencies and coupling gain is 10. This paper also presents numerical examples to investigate the synchronization of the oscillators and convergence to stable equilibrium points.
{"title":"Analytical solutions of equilibrium points of the standard Kuramoto model: 3 and 4 oscillators","authors":"Xin Xin, Takanori Kikkawa, Yannian Liu","doi":"10.1109/ACC.2016.7525284","DOIUrl":"https://doi.org/10.1109/ACC.2016.7525284","url":null,"abstract":"This paper concerns the problem of how to obtain analytically all equilibrium points of the standard Kuramoto model (with all-to-all uniform coupling) having any natural frequencies and coupling gain. As an initial effort to solve this challenging problem, for the Kuramoto model with 3 (respectively 4) oscillators, to obtain analytically all equilibrium points, we show that we need to solve a polynomial equation of the sine of the phase of an oscillator with the highest order of 6 (respectively 14). For 3 oscillators, this polynomial equation with numerical examples shows that the maximal number of distinct equilibrium points for any natural frequencies and coupling gain is 6. For 4 oscillators, this paper shows theoretically that the maximal number of distinct equilibrium points is not greater than 14, and presents two numerical examples to show the existence of 10 distinct equilibrium points. From the numerical investigation carried out in this study, it is a conjecture that the maximal number of distinct equilibrium points of 4 oscillators for all natural frequencies and coupling gain is 10. This paper also presents numerical examples to investigate the synchronization of the oscillators and convergence to stable equilibrium points.","PeriodicalId":137983,"journal":{"name":"2016 American Control Conference (ACC)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127161450","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-07-06DOI: 10.1109/ACC.2016.7526134
D. Katselis, Carolyn L. Beck
The problem of clustering unweighted graphs either in the case of fully or partially observable adjacency matrices is considered in this paper. Both these subproblems have been recently considered in the literature and have been tackled based on convex optimization techniques related to the problems of matrix completion and robust principal component analysis that fall into the general compressive sensing framework. We revisit these approaches and extend them by proposing ways to obtain more accurate clustering results based on better approximations for the ℓ0-norm of a matrix than the ℓ1-norm. The current state-of-art methods correspond to special instances of the proposed extensions. Although nonconvex, the proposed methods can be approximately decomposed into sequences of ℓ1-norm minimization problems, thus pertaining the efficiency of convex formulations. The methods are compared using graphs that are built upon the classical stochastic blockmodel. These comparisons provide a good indication that the proposed methods can improve the accuracy of state-of-art clustering methods applied to currently popular applications such as that of community detection in social networks.
{"title":"Clustering fully and partially observable graphs via nonconvex optimization","authors":"D. Katselis, Carolyn L. Beck","doi":"10.1109/ACC.2016.7526134","DOIUrl":"https://doi.org/10.1109/ACC.2016.7526134","url":null,"abstract":"The problem of clustering unweighted graphs either in the case of fully or partially observable adjacency matrices is considered in this paper. Both these subproblems have been recently considered in the literature and have been tackled based on convex optimization techniques related to the problems of matrix completion and robust principal component analysis that fall into the general compressive sensing framework. We revisit these approaches and extend them by proposing ways to obtain more accurate clustering results based on better approximations for the ℓ0-norm of a matrix than the ℓ1-norm. The current state-of-art methods correspond to special instances of the proposed extensions. Although nonconvex, the proposed methods can be approximately decomposed into sequences of ℓ1-norm minimization problems, thus pertaining the efficiency of convex formulations. The methods are compared using graphs that are built upon the classical stochastic blockmodel. These comparisons provide a good indication that the proposed methods can improve the accuracy of state-of-art clustering methods applied to currently popular applications such as that of community detection in social networks.","PeriodicalId":137983,"journal":{"name":"2016 American Control Conference (ACC)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127166292","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2016-07-06DOI: 10.1109/ACC.2016.7526529
Nicholas Sweet, N. Ahmed
We consider the problem of fusing human-generated semantic `soft sensor' data with conventional `hard sensor' data to augment Bayesian state estimators. This requires modeling semantic soft data via generalized continuous-to-discrete softmax likelihood functions, which can theoretically model semantic descriptions of any dynamic state space. This paper addresses two important related issues for deploying these models in practical applications. First, a general solution to the data-free likelihood synthesis problem is provided. This allows for easy embedding of contextual constraints and other relevant a priori information within generalized softmax models, without resorting to expensive non-convex optimization procedures for parameter estimation with sparse data. This result is then used to derive strategies for combining multiple semantic human observation models into `compressed' likelihood functions for fast batch data fusion. The proposed methods are demonstrated on a human-robot target search application.
{"title":"Structured synthesis and compression of semantic human sensor models for Bayesian estimation","authors":"Nicholas Sweet, N. Ahmed","doi":"10.1109/ACC.2016.7526529","DOIUrl":"https://doi.org/10.1109/ACC.2016.7526529","url":null,"abstract":"We consider the problem of fusing human-generated semantic `soft sensor' data with conventional `hard sensor' data to augment Bayesian state estimators. This requires modeling semantic soft data via generalized continuous-to-discrete softmax likelihood functions, which can theoretically model semantic descriptions of any dynamic state space. This paper addresses two important related issues for deploying these models in practical applications. First, a general solution to the data-free likelihood synthesis problem is provided. This allows for easy embedding of contextual constraints and other relevant a priori information within generalized softmax models, without resorting to expensive non-convex optimization procedures for parameter estimation with sparse data. This result is then used to derive strategies for combining multiple semantic human observation models into `compressed' likelihood functions for fast batch data fusion. The proposed methods are demonstrated on a human-robot target search application.","PeriodicalId":137983,"journal":{"name":"2016 American Control Conference (ACC)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-07-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124876663","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：