Klenk M.,U.S. Navy |
Aha D.W.,U.S. Navy |
Molineaux M.,Knexus Research Corp.
AI Magazine | Year: 2011
Case-based reasoning (CBR) is a problem-solving process in which a new problem is solved by retrieving a similar situation and reusing its solution. CBR can be applied to transfer learning (TL). All CBR methods involve transferring knowledge from prior cases to new problems. In CBR as a transfer learning method, the CBR cycle accounts for all three steps of transfer learning. CBR can be used for problem learning. To be a transfer learning system, the CBR system must be integrated with another component to perform knowledge transfer between the source and target problems. Transfer learning contributes two important metrics for evaluating the flexibility of CBR systems. First, the initial advantage metric empirically measures the flexibility of the CBR system's retrieval and reuse mechanisms. Second, when the system is unable to solve a target problem with the source problems, the learning rate measures the retrieval mechanism's ability to avoid source cases or the CBR system's ability to perform case-based maintenance. Source
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. Source
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: 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.
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.