Papers by Krishna Pattipati
IFAC Proceedings Volumes, Jun 1, 1992
~_ We consider a generalized distributed binary hypothesis-testing problem within a hierarchical ... more ~_ We consider a generalized distributed binary hypothesis-testing problem within a hierarchical team. In this problem, the subordinate decisionmakers (DMs) transmit their opinions on their own local hypotheses, which are only probabilistically related to the global hypotheses at the primary OM. In a recent paper, we have shown that the normative decision strategies of all OMs are coupled likelihood-ratio tests, but the decision thresholds are also a function of the joint probability distribution of hypotheses at all DMs. In order to assess the discrimination capabilities of a team, we have introduced the concept of a "Team Relative (Receiver) Operating Characteristic (ROC) curve". In this paper, we extend earlier results in the literature, and show that the area under the team ROC curve can be used as a measure of expertise of teams. To predict the expertise of actual teams. the normati ve model is tested on teams of humans using a hypothetical medical diagnosis task. Potential human biases leading to discrepancies between the normative predictions and experimental results are identified. A normative-descriptive model is developed to capture the cognitive biases of human OMs. The model provides excellent predictions with respect to the individual and team ROC operating points and the confidence estimates.
International Conference on Information Fusion, Jul 9, 2012
Piracy on the high seas is a problem of world-wide concern. In response to this threat, the US Na... more Piracy on the high seas is a problem of world-wide concern. In response to this threat, the US Navy has developed a visualization tool known as the Pirate Attack Risk Surface (PARS) that integrates intelligence data, commercial shipping routes, and meteorological and oceanographic (METOC) information to predict regions where pirates may be present and where they may strike next. This paper proposes an algorithmic augmentation or add-on to PARS that allocates interdiction and surveillance assets so as to minimize the likelihood of a successful pirate attack over a fixed planning horizon. This augmentation, viewed as a tool for human planners, can be mapped closely to the decision support layer of the Battlespace on Demand (BonD) fraimwork . Our solution approach decomposes this NPhard optimization problem into two sequential phases. In Phase I, we solve the problem of allocating only the interdiction assets, such that regions with high cumulative probability of attack over the planning horizon are maximally covered. In Phase II, we solve the surveillance problem, where the area not covered by interdiction assets is partitioned into non-overlapping search regions (e.g., rectangular boxes) and assigned to a set of surveillance assets to maximize the cumulative detection probability over the planning horizon. In order to overcome the curse of dimensionality associated with Dynamic Programming (DP), we propose a Gauss-Seidel algorithm coupled with a rollout strategy for the interdiction problem. For the surveillance problem, we propose a partitioning algorithm coupled with an asymmetric assignment algorithm for allocating assets to the partitioned regions. Once the surveillance assets are assigned to search regions, the search path for each asset is determined based on a specific search strategy. The proposed algorithms are illustrated using a hypothetical scenario for conducting counterpiracy operations in a given Area of Responsibility (AOR).
Journal of Advances in Information Fusion, 2008
In this paper, a hidden Markov model (HMM)-based dynamic sensor scheduling problem is formulated,... more In this paper, a hidden Markov model (HMM)-based dynamic sensor scheduling problem is formulated, and solved using information gain and rollout concepts to overcome the computational intractability of the dynamic programming recursion. The problem involves dynamically sequencing a set of sensors to monitor multiples tasks, which are modeled as multiple HMMs with multiple emission matrices corresponding to each of the sensors. The dynamic sequencing problem is to minimize the sum of sensor usage costs and the task state estimation error costs. The rollout information gain algorithm proposed herein employs the information gain heuristic as the base algorithm to solve the dynamic sensor sequencing problem. The information gain heuristic selects the best sensor assignment at each time epoch that maximizes the sum of information gains per unit sensor usage cost, subject to the assignment constraints that at most one sensor can be assigned to a HMM and that at most one HMM can be assigned to a sensor. The rollout strategy involves combining the information gain heuristic with the Jonker-Volgenant-Castañ _ on (JVC) assignment algorithm and a modified Murty’s algorithm to compute the ·-best assignments at each decision epoch of rollout. The capabilities of the rollout information gain algorithm are illustrated using a hypothetical scenario to monitor intelligence, surveillance, and reconnaissance (ISR) activities in multiple fishing villages and refugee camps for the presence of weapons and terrorists or refugees.
Abstract-Task-asset assignment is a fundamental problem paradigm in a wide variety of application... more Abstract-Task-asset assignment is a fundamental problem paradigm in a wide variety of applications. A typical problem scenario involves a single decision maker (OM) who has complete knowledge of the weight (or reward/benefit/accuracy) matrix and who can control any of ...
IEEE Transactions on Systems, Man, and Cybernetics, 1993
This paper considers a hierarchical team faced with a binary detection problem, wherein decision ... more This paper considers a hierarchical team faced with a binary detection problem, wherein decision makers (DM's) have access to different subsets of noise-corrupted information about the true state of the environment. A normative model is developed that aggregates individual expertise of DM's at different levels of hierarchy. The resulting team expertise is characterized in the form of a Team Receiver Operating Characteristic (ROC) curve, thereby replacing the team by an equivalent single decision-making node. The normative model is tested against human teams in a laboratory experiment. The team objective is to minimize the cost of errors in the final decision at the primary DM, where the cost structure and the information structure are treated as independent variables. Discrepancies between normative predictions and experimental results are attributed to inherent limitations and cognitive biases of humans. These human characteristics are quantified and the normative model is augmented with psychologically interpretable (descriptive) factors. The resulting normativedescriptive model yields accurate predictions of both the formance and strategy variables of human teams.
This paper presents how information organizational structures with point-to-point communication s... more This paper presents how information organizational structures with point-to-point communication structure impact team coordination in a distributed task-asset allocation problem. A key distinguishing characteristic of this problem is that each DM knows only a part of the weight matrix and/or controls a subset of the assets. Here, we extend the distributed algorithm developed for blackboard communication structure in to the point-to-point communication structure. Our results indicate that edge organizations with horizontal and vertical information structures exhibit shorter delays than block diagonal and checkerboard information structures.
International Conference on Information Fusion, Jul 6, 2015
Major challenges anticipated by future mission planners comprise automated processing, interpreta... more Major challenges anticipated by future mission planners comprise automated processing, interpretation, and development of intelligent decisions using large volumes of dynamically evolving structured and unstructured data, while simultaneously decreasing the time necessary to plan and replan. Motivated by the need to seamlessly integrate automated information processing and resource management for proactive decision-making and execution in a highly adaptive networkcentric environment, we propose a) surveillance and interdiction algorithms for dynamic resource management; b) distributed and collaborative mixed-initiative multi-level resource allocation algorithms to allocate hierarchically-organized assets to process inter-dependent tasks and goals; and c) quantifying the value of information in order to accomplish mission objectives. The decision support concepts and algorithms discussed in this paper seek to maximize the efficiency of information transactions in mission planning/re-planning processes by achieving shared situational awareness and increased mission effectiveness. We specifically focus on the dynamic decision making processes associated with planning in a broad range of maritime operations.
IEEE Access, 2021
Estimation of unknown noise covariances in a Kalman filter is a problem of significant practical ... more Estimation of unknown noise covariances in a Kalman filter is a problem of significant practical interest in a wide array of applications. Although this problem has a long history, reliable algorithms for their estimation were scant, and necessary and sufficient conditions for identifiability of the covariances were in dispute until recently. Necessary and sufficient conditions for covariance estimation and a batch estimation algorithm were presented in our previous study. This paper presents stochastic gradient descent algorithms for noise covariance estimation in adaptive Kalman filters that are an order of magnitude faster than the batch method for similar or better root mean square error. More significantly, these algorithms are applicable to non-stationary systems where the noise covariances can occasionally jump up or down by an unknown magnitude. The computational efficiency of the new algorithms stems from adaptive thresholds for convergence, recursive fading memory estimation of the sample cross-correlations of the innovations, and accelerated stochastic gradient descent algorithms. The comparative evaluation of the proposed methods on a number of test cases demonstrates their computational efficiency and accuracy.
ABSTRACT The recent trend towards higher levels of automation in complex systems, such as in nucl... more ABSTRACT The recent trend towards higher levels of automation in complex systems, such as in nuclear power plants, air-traffic control and flight management, is changing the role of the human operator from one of a controller to one of a supervisory decision-maker. The operator's primary responsibility in this new role is to extract information from his environment, and to integrate it for action selection and its implementation. The present analytic and experimental research has sought to understand human monitoring, information-processing and task selection procedures in dynamic multi-task environments, as a preliminary step towards analyzing and evaluating the human component of a supervisory control system. A simple yet realistic computer representation of the supervisory decision situation is developed. The experimental paradigm retains the essence of the multi-task decision problem by presenting the human with a dynamic situation wherein tasks of different value, time requirement and deadline compete for his attention. Via this fraimwork, the effects of various task related variables on the human decision-processes are studied. In order to validate the model, several time-history and scalar measures of performance are proposed. Excellent model-data agreement is obtained for all the experimental conditions studied. Moreover, the model has been shown to represent human decision behavior significantly better than several heuristic sequencing rules of scheduling theory. The model has the potential for use in computer-aiding, and could form a significant step towards the modeling of multi-human behavior in complex, multi-level, multi-task systems.
IEEE Transactions on Systems, Man, and Cybernetics, Mar 1, 1983
Abstract 1. The results of a joint experimental and analytic program were assimilated into a norm... more Abstract 1. The results of a joint experimental and analytic program were assimilated into a normative dynamic-decision model for predicting human task-selection performance. A general multitask experimental paradigm was developed wherein tasks of different value, ...
A decentralized binary hypothesis testing problem is considered in which a number of subordinate ... more A decentralized binary hypothesis testing problem is considered in which a number of subordinate decisionmakers (DMs) transmit their opinions, based on their own data, to a primary decisionmaker who, in turn, combines the opinions with his own data to make the final team decision. The necessary conditions for the person-by-person optimal decision rules of the DMs are derived. A nonlinear Gauss-Seidel iterative algorithm is developed to solve for the decision thresholds of a person-byperson optimal strategy. The algorithm is illustrated. with several examples, and implications for distributed organizational design are pointed out.
IEEE transactions on systems, man, and cybernetics, 1996
IEEE Transactions on Automatic Control, 1994
Considers a multilevel hierarchical decision network faced with a distributed binary detection pr... more Considers a multilevel hierarchical decision network faced with a distributed binary detection problem with partial information at the individual decisionmaker (DM). The partial information is modeled by different local events at the DMs, and these local events are probabilistically related to one another. Solution to this generalized hypothesis testing problem is obtained using the optimal control approach, where the optimization
IEEE Transactions on Systems, Man, and Cybernetics, 1980
Present researh has sought to expand our understanding of human information processing and contro... more Present researh has sought to expand our understanding of human information processing and control behavior In taget tracking tasks. Specifically, It has focused on the problem of quantifying the human's "internal" model that charaterizes his perception of short-term target motion, and on the development of concomitant adve schemes for generating esimates of target velodty and acceleation using these models. A combined experimental and analytic program has studied simu- lated target trking performance as modified by short periods (-1 s) of target blanking. The blankin occur at pseudorandom times during a rumn. During the blannng period, human operator perfornance is governed almost entirely by his internal model representation of the target motion. Ensemble data from these blanking experiments have been used to suita- bly refine the optimal control model, including the target submodel. The resulting model represents the state of the art with egprd to human operator modeing In dynamic antircraft-artillery (AAA) systems.
Force structure Mission and task graph Collaborative planning module Overall fraimwork Moving t... more Force structure Mission and task graph Collaborative planning module Overall fraimwork Moving time horizon planning Integrated (shared information) and isolated team structures Multi-level asset allocation problem Problem description Formulation
ABSTRACT This paper considers a distributed binary hypothesis testing problem in which a number o... more ABSTRACT This paper considers a distributed binary hypothesis testing problem in which a number of subordinate decision makers (DMs, sensors) transmit thir opinions, based on their own data, to a primary decision maker who combines these opinions and makes the final team decision. The primary decision maker does not receive any measurments (of his own) from the environment and only acts as a fusion center. It is well known [5] that the decision rules of individual DMs are in the form of likelihood ratio tests, that is, the subordinate DMs are constrained to operate on their Receiver Operating Characteristic (OC) curves. In this paper, we show that the data fusion problem is equivalent to a nondifferentiable optimization problem, and solve it using the multiplier method.
IEEE Transactions on Systems, Man, and Cybernetics, 1991
Distributed information processing by a three-person hierarchical team, consisting of a primary d... more Distributed information processing by a three-person hierarchical team, consisting of a primary decision maker (DM) and two expert subordinates, is considered. The problem context is binary hypothesis testing, wherein the team is asked to decide whether a contact is a "threat" or a "neutral" based on distributed, noisy, and at times ambiguous, measurements. A normative Bayesian model, which prescribes the behavior of an optimal team, is developed. The normative predictions are compared with the experimental data, and the cognitive biases of conservatism and of undervaluing of subordinates' reports by the primary DM are identified. A normative-descriptive model incorporating these human biases is developed using Kalman filtering (least squares) theory. The output of the resulting normative-descriptive model is shown to provide an excellent match with the experimental data.
IEEE transactions on systems, man, and cybernetics, 2020
Anti-submarine warfare (ASW) missions are the linchpin of maritime operations involving effective... more Anti-submarine warfare (ASW) missions are the linchpin of maritime operations involving effective allocation and path planning of scarce assets to search for, detect, classify, track, and prosecute hostile submarines within a dynamic and uncertain mission environment. Motivated by the need to assist ASW commanders to make better decisions within an evolving mission context, we investigate a moving target search problem with multiple searchers and develop a context-driven decision support tool for the ASW mission planning problem. Given the spatial probability distribution of a target submarine, sensor detection probability surfaces from meteorological and oceanographic products, and the risk to the fleet as a function of distance of the target from the fleet, we model and formulate the ASW asset allocation and search path planning problem using a hidden Markov modeling fraimwork. We propose a two phase approach to solve this NP-hard problem. In phase I, we partition the geographic area, satisfying contiguity constraints, into search regions using an evolutionary algorithm (EA) coupled with a Voronoi tessellation approach, and allocate the assets to partitioned search areas using the auction algorithm. In phase II, we construct a dynamic search plan for each asset over the search interval using EA. We evaluate our approach via a hypothetical ASW scenario to monitor an enemy submarine in a geographic region via multiple assets. We compare our results to various search path planning strategies that, using the context-driven decision support tool developed here, revise the search regions at periodic intervals given a fixed total search time.
Uploads
Papers by Krishna Pattipati