Auslander B.,Knexus Research Corp. |
Gupta K.M.,Knexus Research Corp. |
Aha D.W.,U.S. Navy
Proceedings of SPIE - The International Society for Optical Engineering | Year: 2011
A variety of anomaly detection algorithms have been applied to surveillance tasks for detecting threats with some success. However, it is not clear which anomaly detection algorithms should be used for domains such as ground-based maritime video surveillance. For example, recently introduced algorithms that use local density techniques have performed well for some tasks, but they have not been applied to ground-based maritime video surveillance. Also, the reasons for the performance differences of anomaly detection algorithms on problems of varying difficulty are not well understood. We address these two issues by comparing families of global and local anomaly detection algorithms on tracks extracted from ground-based maritime surveillance videos. Obtaining maritime anomaly data can be difficult or even impractical. Therefore, we use a generative approach to vary and control the difficulty of anomaly detection tasks and to focus on borderline and difficult situations in our empirical comparison studies. We report that global algorithms outperform local algorithms when tracks have large and unstructured variations, while they perform equally well when the tracks have only minor variations. © 2011 SPIE.
Klenk M.,U.S. Navy |
Molineaux M.,Knexus Research Corp. |
Aha D.W.,U.S. Navy
Computational Intelligence | Year: 2013
To operate autonomously in complex environments, an agent must monitor its environment and determine how to respond to new situations. To be considered intelligent, an agent should select actions in pursuit of its goals, and adapt accordingly when its goals need revision. However, most agents assume that their goals are given to them; they cannot recognize when their goals should change. Thus, they have difficulty coping with the complex environments of strategy simulations that are continuous, partially observable, dynamic, and open with respect to new objects. To increase intelligent agent autonomy, we are investigating a conceptual model for goal reasoning called Goal-Driven Autonomy (GDA), which allows agents to generate and reason about their goals in response to environment changes. Our hypothesis is that GDA enables an agent to respond more effectively to unexpected events in complex environments. We instantiate the GDA model in ARTUE (Autonomous Response to Unexpected Events), a domain-independent autonomous agent. We evaluate ARTUE on scenarios from two complex strategy simulations, and report on its comparative benefits and limitations. By employing goal reasoning, ARTUE outperforms an off-line planner and a discrepancy-based replanner on scenarios requiring reasoning about unobserved objects and facts and on scenarios presenting opportunities outside the scope of its current mission. © 2012 Wiley Periodicals, Inc.
Floyd M.W.,Knexus Research Corp. |
Drinkwater M.,Knexus Research Corp. |
Aha D.W.,U.S. Navy
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) | Year: 2014
Robots are added to human teams to increase the team's skills or capabilities. To gain the acceptance of the human teammates, it may be important for the robot to behave in a manner that the teammates consider trustworthy. We present an approach that allows a robot's behavior to be adapted so that it behaves in a trustworthy manner. The adaptation is guided by an inverse trust metric that the robot uses to estimate the trust a human teammate has in it. We evaluate our method in a simulated robotics domains and demonstrate how the agent can adapt to a teammate's preferences. © 2014 Springer International Publishing.
Knexus Research Corp. | Date: 2015-11-03
A computer system and method according to the present invention can receive multi-modal inputs such as natural language, gesture, text, sketch and other inputs in order to simplify and improve locative question answering in virtual worlds, among other tasks. The components of an agent as provided in accordance with one embodiment of the present invention can include one or more sensors, actuators, and cognition elements, such as interpreters, executive function elements, working memory, long term memory and reasoners for responses to locative queries, for example. Further, the present invention provides, in part, a locative question answering algorithm, along with the command structure, vocabulary, and the dialog that an agent is designed to support in accordance with various embodiments of the present invention.
Agency: Department of Defense | Branch: Navy | Program: SBIR | Phase: Phase I | Award Amount: 149.97K | Year: 2012
Mine Countermeasures (MCM) missions of tomorrow will increasingly exploit autonomous vehicles such as the remote multi-mission vehicle to reduce risk to personnel and equipment. However, existing approaches are limited for continuous sensing and planning needed to deal with the dynamic and uncertain nature of the MCM missions. To address this gap we will consider alternative approaches that merge symbolic and probabilistic reasoning, a complementary approach to existing mathematical optimization, to handle planning and replanning under uncertainty. We will develop and investigate a system called CAMPER (A Cognitive Architecture for MCM planning, execution, and replanning). CAMPER will respond effectively to evolving MCM environments by making and executing plans that coordinate multiple autonomous vehicles to achieve goals. This will form the basis of a promising operator friendly system that addresses the problem of decision support and automation in high-risk, high-uncertainty MCM missions. As part of CAMPER, we will investigate novel methods for deliberative planning under uncertainty, environment monitoring using distributed sensors, and explanation as a means of understanding the environment and engendering trust. We will demonstrate the feasibility of CAMPER by evaluating it on simulated missions.
Agency: Department of Defense | Branch: Navy | Program: SBIR | Phase: Phase II | Award Amount: 733.53K | Year: 2011
Currently, there are no robust software platforms available for developing and evaluating reusable, virtual, communicative spatio-temporal agents. This makes rapid development and meaningful comparative evaluation, critical requirements for military and non-military applications, infeasible. To meet this need, we will develop a software platform called CoASTeR-WB (Communicative Agent for Spatio-Temporal Reasoning), and supporting Knowledge Engineering tools, and utilities. We will build on CoASTeR-WB Phase I architecture to develop a mature and advanced version that introduces additional human-computer interaction modalities such as sketch, speech, and gesture using open source and COTS components. We will research and develop advanced versions of perception, spatio-temporal reasoning, and action planning components using cognitive architectures and state-of-the-art AI planning techniques. We will fully develop CoASTeR-WB"s automated test and evaluation framework for assessing the performance of communicative dynamic agents in navigation and extra navigational tasks. We will evaluate CoASTeR-WB with navy relevant scenarios of unmanned systems missions using the Workbench"s own automated evaluation framework. We will develop knowledge engineering and test and evaluation authoring tools to make CoASTeR-WB more accessible to developers and researchers.
Agency: Department of Defense | Branch: Navy | Program: SBIR | Phase: Phase II | Award Amount: 730.27K | Year: 2013
Mine countermeasures (MCM) missions require complex coordination of personnel, equipment, and autonomous vehicles. The mission success hinges on timely and accurate decision-making for responding to dynamic situations in the area of operations that pose considerable risks to mission assets; both personnel and equipment. However, no decision support tools are presently available for mission personnel to exploit the most up to date situation information and maximize mission performance. To fill this capability gap, we will develop a decision support system called CARPE; a Cognitive Architecture for RePlanning and Execution, that shall support decision making for MCM missions with replanning, retasking, and rescheduling capabilities, and shall interoperate with the MEDAL-EA system. We will develop novel techniques for deviation detection, interactive deviation cause analysis, minimally disruptive rescheduling, causal and plan model learning. We will implement and integrate these technologies as reasoning services with MEDAL-EA to provide a seamless decision support experience to MCM mission personnel. We will deliver progressively mature versions of CARPE that will be evaluated by Subject Matter Experts using simulation assisted decision-making.
Agency: Department of Defense | Branch: Navy | Program: SBIR | Phase: Phase I | Award Amount: 149.97K | Year: 2013
Unmanned systems have proven their value in combat operations by delivering unprecedented mission performance. Unfortunately, the current unmanned systems are predominantly tele-operated, which tie up many skilled operators per unmanned system. Thus, increasing demands for unmanned system cannot be met with the current state-of-the-art. The problem of low levels of autonomy is further exacerbated by the lack of decision support for behavioral anomaly detection and subsequent recovery planning. We will address this capability gap by developing approaches for increasing the level of autonomy from tele-operated to human supervised as follows. In particular, we will develop ADRUS, a decision support system for anomaly detection and recovery for unmanned system for multi-vehicle missions. ADRUS will provide automated monitoring, perform continuous anomaly detection and analysis in the mission context, analyze root causes for the anomaly and explain its findings to the mission personnel. It will go a step further and recommend plans to recover from the anomaly to minimize disruptions and maximize mission success. To develop these capabilities, we will investigate the use of a variety of probabilistic causal models that exploit the knowledge of mission to assess the deviations and provide accurate alerts. We will investigate fast and incremental automated planning approaches that exploit current resource knowledge to compute effective recovery plans. In developing ADRUS, we will consider human factor issues, such as reduction of cognitive load by developing appropriate alert presentation techniques and human-machine trust by developing decision explanation and justification abilities. We will demonstrate ADRUS feasibility by developing a prototype reference implementation and evaluating it using multi-vehicle mission scenarios.
Agency: Department of Defense | Branch: Navy | Program: SBIR | Phase: Phase I | Award Amount: 99.49K | Year: 2010
Currently there are no software platforms for developing and evaluating reusable, virtual, communicative spatio-temporal agents. This makes rapid development and meaningful comparative evaluation, critical requirements for military and non-military applications, infeasible. Lack of such a platform also slows down the much needed research in spatio-temporal technologies. We will architect a software workbench called CoASTeR-WB (Communicative Agent for Spatio-Temporal Reasoning) to meet these requirements. In contrast to the state-of-the-art technologies that use disparate and proprietary representation and reasoning technologies, we will investigate a unique framework for plug-and-play open-source virtual worlds and communicative technologies and reusable spatio-temporal reasoning agents. We will investigate automatic methods for achieving semantic interoperability among CoASTeR-WB components. We will investigate the use of a cognitive architecture as a robust foundation for developing spatio-temporal agents for advanced navigation and extra-navigational tasks. Finally, we will develop specifications for a built-in evaluation framework with a library of spatio-temporal reasoning tasks and annotated scenarios to enable rapid, consistent, and comparative evaluations. We will validate our approach by implementing a proof-of-concept prototype and executing an evaluation run with a library of sample spatio-temporal reasoning test-problems.
Agency: Department of Defense | Branch: Office of the Secretary of Defense | Program: SBIR | Phase: Phase II | Award Amount: 999.71K | Year: 2014
Deployment of unmanned systems continues to expand across a wide range of missions; for example, logistics and resupply missions, force application and protection, and improving battlespace awareness. Presently these unmanned systems run at the lowest of four possible levels of autonomy, that is, in a teleoperated mode, and each system typically requires multiple operators. To address this problem of operator scalability, investigations into approaches for human supervised autonomy were called for in this SBIR. In Phase I, we took a step toward addressing this capability gap by developing a decision support system for Anomaly Detection and Recovery of Unmanned Systems (ADRUS). In particular, we demonstrated that ADRUS could successfully handle unexpected events or anomalies and replan to recover from them. Our demonstration included a proof-of-concept prototype implementation and its performance in simulated logistics and resupply missions. In Phase II, we will continue algorithmic development of anomaly detection, mission risk analysis, and replanning reasoning services to meet the performance requirements of our target transition environments. Our approaches and extensions will include methods for improving reasoning accuracies, model coverage and fidelity, as well as the ability to learn and improve knowledge models by exploiting operator interactions and decisions data. We will implement and evaluate progressively mature versions of ADRUS throughout the performance period. We will conduct repeated tests and evaluations (T&E) in simulation using realistic models of target unmanned platforms. Based on T&E, we will characterize the robustness, scalability, and coverage of ADRUS. In addition, we will evaluate the operational effectiveness resulting from human supervisory control enabled by ADRUS. For these evaluations, we will engage application subject matter experts (SME) and operators from candidate transition programs. We have initiated discussions with prime performers from target programs developing these unmanned platforms and we will develop our transition requirements accordingly.