The purpose of this LDRD is to develop technology allowing warfighters to provide high-level commands to their unmanned assets, freeing them to command a group of them or commit the bulk of their attention elsewhere. To this end, a brain-emulating cognition and control architecture (BECCA) was developed, incorporating novel and uniquely capable feature creation and reinforcement learning algorithms. BECCA was demonstrated on both a mobile manipulator platform and on a seven degree of freedom serial link robot arm. Existing military ground robots are almost universally teleoperated and occupy the complete attention of an operator. They may remove a soldier from harm's way, but they do not necessarily reduce manpower requirements. Current research efforts to solve the problem of autonomous operation in an unstructured, dynamic environment fall short of the desired performance. In order to increase the effectiveness of unmanned vehicle (UV) operators, we proposed to develop robots that can be 'directed' rather than remote-controlled. They are instructed and trained by human operators, rather than driven. The technical approach is modeled closely on psychological and neuroscientific models of human learning. Two Sandia-developed models are utilized in this effort: the Sandia Cognitive Framework (SCF), a cognitive psychology-based model of human processes, and BECCA, a psychophysical-based model of learning, motor control, and conceptualization. Together, these models span the functional space from perceptuo-motor abilities, to high-level motivational and attentional processes.
The neocortex is perhaps the highest region of the human brain, where audio and visual perception takes place along with many important cognitive functions. An important research goal is to describe the mechanisms implemented by the neocortex. There is an apparent regularity in the structure of the neocortex [Brodmann 1909, Mountcastle 1957] which may help simplify this task. The work reported here addresses the problem of how to describe the putative repeated units ('cortical circuits') in a manner that is easily understood and manipulated, with the long-term goal of developing a mathematical and algorithmic description of their function. The approach is to reduce each algorithm to an enhanced perceptron-like structure and describe its computation using difference equations. We organize this algorithmic processing into larger structures based on physiological observations, and implement key modeling concepts in software which runs on parallel computing hardware.
This 3-year research and development effort focused on what we believe is a significant technical gap in existing modeling and simulation capabilities: the representation of plausible human cognition and behaviors within a dynamic, simulated environment. Specifically, the intent of the ''Simulating Human Behavior for National Security Human Interactions'' project was to demonstrate initial simulated human modeling capability that realistically represents intra- and inter-group interaction behaviors between simulated humans and human-controlled avatars as they respond to their environment. Significant process was made towards simulating human behaviors through the development of a framework that produces realistic characteristics and movement. The simulated humans were created from models designed to be psychologically plausible by being based on robust psychological research and theory. Progress was also made towards enhancing Sandia National Laboratories existing cognitive models to support culturally plausible behaviors that are important in representing group interactions. These models were implemented in the modular, interoperable, and commercially supported Umbra{reg_sign} simulation framework.
This report describes work carried out under a Sandia National Laboratories Excellence in Engineering Fellowship in the Department of Electrical and Computer Engineering at the University of Illinois at Urbana-Champaign. Our research group (at UIUC) is developing a intelligent robot, and attempting to teach it language. While there are many aspects of this research, for the purposes of this report the most important are the following ideas. Language is primarily based on semantics, not syntax. To truly learn meaning, the language engine must be part of an embodied intelligent system, one capable of using associative learning to form concepts from the perception of experiences in the world, and further capable of manipulating those concepts symbolically. In the work described here, we explore the use of hidden Markov models (HMMs) in this capacity. HMMs are capable of automatically learning and extracting the underlying structure of continuous-valued inputs and representing that structure in the states of the model. These states can then be treated as symbolic representations of the inputs. We describe a composite model consisting of a cascade of HMMs that can be embedded in a small mobile robot and used to learn correlations among sensory inputs to create symbolic concepts. These symbols can then be manipulated linguistically and used for decision making. This is the project final report for the University Collaboration LDRD project, 'A Robotic Framework for Semantic Concept Learning'.
This report documents work undertaken to endow the cognitive framework currently under development at Sandia National Laboratories with a human-like memory for specific life episodes. Capabilities have been demonstrated within the context of three separate problem areas. The first year of the project developed a capability whereby simulated robots were able to utilize a record of shared experience to perform surveillance of a building to detect a source of smoke. The second year focused on simulations of social interactions providing a queriable record of interactions such that a time series of events could be constructed and reconstructed. The third year addressed tools to promote desktop productivity, creating a capability to query episodic logs in real time allowing the model of a user to build on itself based on observations of the user's behavior.
This project is being conducted by Sandia National Laboratories in support of the DARPA Augmented Cognition program. Work commenced in April of 2002. The objective for the DARPA program is to 'extend, by an order of magnitude or more, the information management capacity of the human-computer warfighter.' Initially, emphasis has been placed on detection of an operator's cognitive state so that systems may adapt accordingly (e.g., adjust information throughput to the operator in response to workload). Work conducted by Sandia focuses on development of technologies to infer an operator's ongoing cognitive processes, with specific emphasis on detecting discrepancies between machine state and an operator's ongoing interpretation of events.
The Geometric Search Engine is a software system for storing and searching a database of geometric models. The database maybe searched for modeled objects similar in shape to a target model supplied by the user. The database models are generally from CAD models while the target model may be either a CAD model or a model generated from range data collected from a physical object. This document describes key generation, database layout, and search of the database.
Umbra is a new Sandia-developed modeling and simulation framework. The Umbra framework allows users to quickly build models and simulations for intelligent system development, analysis, experimentation, and control and supports tradeoff analyses of complex robotic systems, device, and component concepts. Umbra links together heterogeneous collections of modeling tools. The models in Umbra include 3D geometry and physics models of robots, devices and their environments. Model components can be built with varying levels of fidelity and readily switched to allow models built with low fidelity for conceptual analysis to be gradually converted to high fidelity models for later phase detailed analysis. Within control environments, the models can be readily replaced with actual control elements. This paper describes Umbra at a functional level and describes issues that Sandia uses Umbra to address.
We consider the problem of planning shortest paths for a tethered robot with a finite length tether in a 2D environment with polygonal obstacles. We present an algorithm that runs in time O((k{sub 1} + 1){sup 2}n{sup 4}) and finds the shortest path or correctly determines that none exists that obeys the constraints; here n is the number obstacle vertices, and k{sub 1} is the number loops in the initial configuration of the tether. The robot may cross its tether but nothing can cross obstacles, which cause the tether to bend. The algorithm applies as well for planning a shortest path for the free end of an anchored cable.
One role of the Intelligent Robotics and System Center (ISRC) at Sandia National Laboratories is to address certain aspects of Sandia`s mission to design, manufacture, maintain, and dismantle nuclear weapon components. Hazardous materials, devices, and environments are often involved. Because of shrinking resources, these tasks must be accomplished with a minimum of prototyping, while maintaining high reliability. In this paper, the authors describe simulation, off-line programming/planning, and related tools which are in use, under development, and being researched to solve these problems at the ISRC.
The goal of this project was to develop a framework for robotic planning and execution that provides a continuum of adaptability with respect to model incompleteness, model error, and sensing error. For example, dividing robot motion into gross-motion planning, fine-motion planning, and sensor-augmented control had yielded productive research and solutions to individual problems. Unfortunately, these techniques could only be combined by hand with ad hoc methods and were restricted to systems where all kinematics are completely modeled in planning. The original intent was to develop methods for understanding and autonomously synthesizing plans that coordinate motion, sensing, and control. The project considered this problem from several perspectives. Results included (1) theoretical methods to combine and extend gross-motion and fine-motion planning; (2) preliminary work in flexible-object manipulation and an implementable algorithm for planning shortest paths through obstacles for the free-end of an anchored cable; (3) development and implementation of a fast swept-body distance algorithm; and (4) integration of Sandia`s C-Space Toolkit geometry engine and SANDROS motion planer and improvements, which yielded a system practical for everyday motion planning, with path-segment planning at interactive speeds. Results (3) and (4) have either led to follow-on work or are being used in current projects, and they believe that (2) will eventually be also.
A robot`s configuration space (c-space) is the space of its kinematic degrees of freedom, e.g., the joint-space of an arm. Sets in c-space can be defined that characterize a variety of spatial relationships, such as contact between the robot and its environment. C-space techniques have been fundamental to research progress in areas such as motion planning and physically-based reasoning. However, practical progress has been slowed by the difficulty of implementing the c-space abstraction inside each application. For this reason, we proposed a Configuration Space Toolkit of high-performance algorithms and data structures meeting these needs. Our intent was to develop this robotics software to provide enabling technology to emerging applications that apply the c-space abstraction, such as advanced motion planning, teleoperation supervision, mechanism functional analysis, and design tools. This final report presents the research results and technical achievements of this LDRD project. Key results and achievements included (1) a hybrid Common LISP/C prototype that implements the basic C-Space abstraction, (2) a new, generic, algorithm for constructing hierarchical geometric representations, and (3) a C++ implementation of an algorithm for fast distance computation, interference detection, and c-space point-classification. Since the project conclusion, motion planning researchers in Sandia`s Intelligent Systems and Robotics Center have been using the CSTk libcstk.so C++ library. The code continues to be used, supported, and improved by projects in the ISRC.
For a number of years, robotics researchers have exploited hierarchical representations of geometrical objects and scenes in motion-planning, collision-avoidance, and simulation. However, few general techniques exist for automatically constructing them. We present a generic, bottom-up algorithm that uses a heuristic clustering technique to produced balanced, coherent hierarchies. Its worst-case running time is O(N{sup 2}logN), but for non-pathological cases it is O(NlogN), where N is the number of input primitives. We have completed a preliminary C++ implementation for input collections of 3D convex polygons and 3D convex polyhedra and conducted simple experiments with scenes of up to 12,000 polygons, which take only a few minutes to process. We present examples using spheres and convex hulls as hierarchy primitives.