Organizational Unit:

Institute for Robotics and Intelligent Machines (IRIM)

Permanent Link

https://hdl.handle.net/1853/70657

Includes Organization(s)

Organizational Unit

Biorobotics and Human Modeling Lab

Organizational Unit

Humanoid Robotics Laboratory

Full item page

Publication Search Results

Now showing 1 - 10 of 13

The role of trust and relationships in human-robot social interaction

(Georgia Institute of Technology, 2009-11-10) Wagner, Alan Richard

Can a robot understand a human's social behavior? Moreover, how should a robot act in response to a human's behavior? If the goals of artificial intelligence are to understand, imitate, and interact with human level intelligence then researchers must also explore the social underpinnings of this intellect. Our endeavor is buttressed by work in biology, neuroscience, social psychology and sociology. Initially developed by Kelley and Thibaut, social psychology's interdependence theory serves as a conceptual skeleton for the study of social situations, a computational process of social deliberation, and relationships (Kelley&Thibaut, 1978). We extend and expand their original work to explore the challenge of interaction with an embodied, situated robot. This dissertation investigates the use of outcome matrices as a means for computationally representing a robot's interactions. We develop algorithms that allow a robot to create these outcome matrices from perceptual information and then to use them to reason about the characteristics of their interactive partner. This work goes on to introduce algorithms that afford a means for reasoning about a robot's relationships and the trustworthiness of a robot's partners. Overall, this dissertation embodies a general, principled approach to human-robot interaction which results in a novel and scientifically meaningful approach to topics such as trust and relationships.
Visual place categorization

(Georgia Institute of Technology, 2009-07-06) Wu, Jianxin

Knowing the semantic category of a robot's current position not only facilitates the robot's navigation, but also greatly improves its ability to serve human needs and to interpret the scene. Visual Place Categorization (VPC) is addressed in this dissertation, which refers to the problem of predicting the semantic category of a place using visual information collected from an autonomous robot platform. Census Transform (CT) histogram and Histogram Intersection Kernel (HIK) based visual codebooks are proposed to represent an image. CT histogram encodes the stable spatial structure of an image that reflects the functionality of a location. It is suitable for categorizing places and has shown better performance than commonly used descriptors such as SIFT or Gist in the VPC task. HIK has been shown to work better than the Euclidean distance in classifying histograms. We extend it in an unsupervised manner to generate visual codebooks for the CT histogram descriptor. HIK codebooks help CT histogram to deal with the huge variations in VPC and improve system accuracy. A computational method is also proposed to generate HIK codebooks in an efficient way. The first significant VPC dataset in home environments is collected and is made publicly available, which is also used to evaluate the VPC system based on the proposed techniques. The VPC system achieves promising results for this challenging problem, especially for important categories such as bedroom, bathroom, and kitchen. The proposed techniques achieved higher accuracies than competing descriptors and visual codebook generation methods.
Automatic coordination and deployment of multi-robot systems

(Georgia Institute of Technology, 2009-03-31) Smith, Brian Stephen

We present automatic tools for configuring and deploying multi-robot networks of decentralized, mobile robots. These methods are tailored to the decentralized nature of the multi-robot network and the limited information available to each robot. We present methods for determining if user-defined network tasks are feasible or infeasible for the network, considering the limited range of its sensors. To this end, we define rigid and persistent feasibility and present necessary and sufficient conditions (along with corresponding algorithms) for determining the feasibility of arbitrary, user-defined deployments. Control laws for moving multi-robot networks in acyclic, persistent formations are defined. We also present novel Embedded Graph Grammar Systems (EGGs) for coordinating and deploying the network. These methods exploit graph representations of the network, as well as graph-based rules that dictate how robots coordinate their control. Automatic systems are defined that allow the robots to assemble arbitrary, user-defined formations without any reliance on localization. Further, this system is augmented to deploy these formations at the user-defined, global location in the environment, despite limited localization of the network. The culmination of this research is an intuitive software program with a Graphical User Interface (GUI) and a satellite image map which allows users to enter the desired locations of sensors. The automatic tools presented here automatically configure an actual multi-robot network to deploy and execute user-defined network tasks.
Incremental smoothing and mapping

(Georgia Institute of Technology, 2008-11-17) Kaess, Michael

Incremental smoothing and mapping (iSAM) is presented, a novel approach to the simultaneous localization and mapping (SLAM) problem. SLAM is the problem of estimating an observer's position from local measurements only, while creating a consistent map of the environment. The problem is difficult because even very small errors in the local measurements accumulate over time and lead to large global errors. iSAM provides an exact and efficient solution to the SLAM estimation problem while also addressing data association. For the estimation problem, iSAM provides an exact solution by performing smoothing, which keeps all previous poses as part of the estimation problem, and therefore avoids linearization errors. iSAM uses methods from sparse linear algebra to provide an efficient incremental solution. In particular, iSAM deploys a direct equation solver based on QR matrix factorization of the naturally sparse smoothing information matrix. Instead of refactoring the matrix whenever new measurements arrive, only the entries of the factor matrix that actually change are calculated. iSAM is efficient even for robot trajectories with many loops as it performs periodic variable reordering to avoid unnecessary fill-in in the factor matrix. For the data association problem, I present state of the art data association techniques in the context of iSAM and present an efficient algorithm to obtain the necessary estimation uncertainties in real-time based on the factored information matrix. I systematically evaluate the components of iSAM as well as the overall algorithm using various simulated and real-world data sets.
Object categorization for affordance prediction

(Georgia Institute of Technology, 2008-07-01) Sun, Jie

A fundamental requirement of any autonomous robot system is the ability to predict the affordances of its environment, which define how the robot can interact with various objects. In this dissertation, we demonstrate that the conventional direct perception approach can indeed be applied to the task of training robots to predict affordances, but it does not consider that objects can be grouped into categories such that objects of the same category have similar affordances. Although the connection between object categorization and the ability to make predictions of attributes has been extensively studied in cognitive science research, it has not been systematically applied to robotics in learning to predict a number of affordances from recognizing object categories. We develop a computational framework of learning and predicting affordances where a robot explicitly learns the categories of objects present in its environment in a partially supervised manner, and then conducts experiments to interact with the objects to both refine its model of categories and the category-affordance relationships. In comparison to the direct perception approach, we demonstrate that categories make the affordance learning problem scalable, in that they make more effective use of scarce training data and support efficient incremental learning of new affordance concepts. Another key aspect of our approach is to leverage the ability of a robot to perform experiments on its environment and thus gather information independent of a human trainer. We develop the theoretical underpinnings of category-based affordance learning and validate our theory on experiments with physically-situated robots. Finally, we refocus the object categorization problem of computer vision back to the theme of autonomous agents interacting with a physical world consisting of categories of objects. This enables us to reinterpret and extend the Gluck-Corter category utility function for the task of learning categorizations for affordance prediction.
Dynamics of embodied dissociated cortical cultures for the control of hybrid biological robots.

(Georgia Institute of Technology, 2007-11-14) Bakkum, Douglas James

The thesis presents a new paradigm for studying the importance of interactions between an organism and its environment using a combination of biology and technology: embodying cultured cortical neurons via robotics. From this platform, explanations of the emergent neural network properties leading to cognition are sought through detailed electrical observation of neural activity. By growing the networks of neurons and glia over multi-electrode arrays (MEA), which can be used to both stimulate and record the activity of multiple neurons in parallel over months, a long-term real-time 2-way communication with the neural network becomes possible. A better understanding of the processes leading to biological cognition can, in turn, facilitate progress in understanding neural pathologies, designing neural prosthetics, and creating fundamentally different types of artificial cognition. Here, methods were first developed to reliably induce and detect neural plasticity using MEAs. This knowledge was then applied to construct sensory-motor mappings and training algorithms that produced adaptive goal-directed behavior. To paraphrase the results, most any stimulation could induce neural plasticity, while the inclusion of temporal and/or spatial information about neural activity was needed to identify plasticity. Interestingly, the plasticity of action potential propagation in axons was observed. This is a notion counter to the dominant theories of neural plasticity that focus on synaptic efficacies and is suggestive of a vast and novel computational mechanism for learning and memory in the brain. Adaptive goal-directed behavior was achieved by using patterned training stimuli, contingent on behavioral performance, to sculpt the network into behaviorally appropriate functional states: network plasticity was not only induced, but could be customized. Clinically, understanding the relationships between electrical stimulation, neural activity, and the functional expression of neural plasticity could assist neuro-rehabilitation and the design of neuroprosthetics. In a broader context, the networks were also embodied with a robotic drawing machine exhibited in galleries throughout the world. This provided a forum to educate the public and critically discuss neuroscience, robotics, neural interfaces, cybernetics, bio-art, and the ethics of biotechnology.
Acoustical Awareness for Intelligent Robotic Action

(Georgia Institute of Technology, 2007-11-12) Martinson, Eric Beowulf

With the growth of successes in pattern recognition and signal processing, mobile robot applications today are increasingly equipping their hardware with microphones to improve the set of available sensory information. However, if the robot, and therefore the microphone, ends up in a poor location acoustically, then the data will remain noisy and potentially useless for accomplishing the required task. This is compounded by the fact that there are many bad acoustic locations through which a robot is likely to pass, and so the results from auditory sensors often remain poor for much of the task. The movement of the robot, though, can also be an important tool for overcoming these problems, a tool that has not been exploited in the traditional signal processing community. Robots are not limited to a single location as are traditionally placed microphones, nor are they powerless over to where they will be moved as with wearable computers. If there is a better location available for performing its task, a robot can navigate to that location under its own power. Furthermore, when deciding where to move, robots can develop complex models of the environment. Using an array of sensors, a mobile robot can build models of sound flow through an area, picking from those models the paths most likely to improve performance of an acoustic application. In this dissertation, we address the question of how to exploit robotic movement. Using common sensors, we present a collection of tools for gathering information about the auditory scene and incorporating that information into a general framework for acoustical awareness. Thus equipped, robots can make intelligent decisions regarding control strategies to enhance their performance on the underlying acoustic application.
Adaptive Estimation and Control with Application to Vision-based Autonomous Formation Flight

(Georgia Institute of Technology, 2007-05-17) Sattigeri, Ramachandra Jayant

The role of vision as an additional sensing mechanism has received a lot of attention in recent years in the context of autonomous flight applications. Modern Unmanned Aerial Vehicles (UAVs) are equipped with vision sensors because of their light-weight, low-cost characteristics and also their ability to provide a rich variety of information of the environment in which the UAVs are navigating in. The problem of vision based autonomous flight is very difficult and challenging since it requires bringing together concepts from image processing and computer vision, target tracking and state estimation, and flight guidance and control. This thesis focuses on the adaptive state estimation, guidance and control problems involved in vision-based formation flight. Specifically, the thesis presents a composite adaptation approach to the partial state estimation of a class of nonlinear systems with unmodeled dynamics. In this approach, a linear time-varying Kalman filter is the nominal state estimator which is augmented by the output of an adaptive neural network (NN) that is trained with two error signals. The benefit of the proposed approach is in its faster and more accurate adaptation to the modeling errors over a conventional approach. The thesis also presents two approaches to the design of adaptive guidance and control (G&C) laws for line-of-sight formation flight. In the first approach, the guidance and autopilot systems are designed separately and then combined together by assuming time-scale separation. The second approach is based on integrating the guidance and autopilot design process. The developed G&C laws using both approaches are adaptive to unmodeled leader aircraft acceleration and to own aircraft aerodynamic uncertainties. The thesis also presents theoretical justification based on Lyapunov-like stability analysis for integrating the adaptive state estimation and adaptive G&C designs. All the developed designs are validated in nonlinear, 6DOF fixed-wing aircraft simulations. Finally, the thesis presents a decentralized coordination strategy for vision-based multiple-aircraft formation control. In this approach, each aircraft in formation regulates range from up to two nearest neighboring aircraft while simultaneously tracking nominal desired trajectories common to all aircraft and avoiding static obstacles.
Optimal, Multi-Modal Control with Applications in Robotics

(Georgia Institute of Technology, 2007-04-04) Mehta, Tejas R.

The objective of this dissertation is to incorporate the concept of optimality to multi-modal control and apply the theoretical results to obtain successful navigation strategies for autonomous mobile robots. The main idea in multi-modal control is to breakup a complex control task into simpler tasks. In particular, number of control modes are constructed, each with respect to a particular task, and these modes are combined according to some supervisory control logic in order to complete the overall control task. This way of modularizing the control task lends itself particularly well to the control of autonomous mobile robot, as evidenced by the success of behavior-based robotics. Many challenging and interesting research issues arise when employing multi-modal control. This thesis aims to address these issues within an optimal control framework. In particular, the contributions of this dissertation are as follows: We first addressed the problem of inferring global behaviors from a collection of local rules (i.e., feedback control laws). Next, we addressed the issue of adaptively varying the multi-modal control system to further improve performance. Inspired by adaptive multi-modal control, we presented a constructivist framework for the learning from example problem. This framework was applied to the DARPA sponsored Learning Applied to Ground Robots (LAGR) project. Next, we addressed the optimal control of multi-modal systems with infinite dimensional constraints. These constraints are formulated as multi-modal, multi-dimensional (M3D) systems, where the dimensions of the state and control spaces change between modes to account for the constraints, to ease the computational burdens associated with traditional methods. Finally, we used multi-modal control strategies to develop effective navigation strategies for autonomous mobile robots. The theoretical results presented in this thesis are verified by conducting simulated experiments using Matlab and actual experiments using the Magellan Pro robot platform and the LAGR robot. In closing, the main strength of multi-modal control lies in breaking up complex control task into simpler tasks. This divide-and-conquer approach helps modularize the control system. This has the same effect on complex control systems that object-oriented programming has for large-scale computer programs, namely it allows greater simplicity, flexibility, and adaptability.
Graph-based Path Planning for Mobile Robots

(Georgia Institute of Technology, 2006-11-16) Wooden, David T.

In this thesis, questions of navigation, planning and control of real-world mobile robotic systems are addressed. Chapter II contains the first contribution in this thesis, which is a modification of the canonical two-layer hybrid architecture: deliberative planning on top, with reactive behaviors underneath. Deliberative is used to describe higher-level reasoning that includes experiential memory and regional or global objectives. Alternatively, reactive describes low-level controllers that operate on information spatially and temporally immediate to the robot. In the traditional architecture, information is passed top down, with the deliberative layer dictating to the reactive layer. Chapter II presents our work on introducing feedback in the opposite direction, allowing the behaviors to provide information to the planning module(s). The path planning problem, particularly as it as solved by the visibility graph, is addressed first in Chapter III. Our so-called oriented visibility graph is a combinatorial planner with emphasis on dynamic re-planning in unknown environments at the expensive of guaranteed optimality at all times. An example of single source planning -- where the goal location is known and static -- this approach is compared to related approaches (e.g. the reduced visibility graph). The fourth chapter further develops the work presented in the Chapter III; the oriented visibility graph is extended to the hierarchical oriented visibility graph. This work directly addresses some of the limitations of the oriented visibility graph, particularly the loss of optimality in the case where obstacles are non-convex and where the convex hulls of obstacles overlap. This results in an approach that is a kind of middle-ground between the oriented visibility graph which was designed to handle dynamic updates very fast, and the reduced visibility graph, an old standard in path planning that guarantees optimality. Chapter V investigates path planning at a higher level of abstraction. Given is a weighted colored graph where vertices are assigned a color (or class) that indicates a feature or quality of the environment associated with that vertex. The question is then asked, ``what is the globally optimal path through this weighted colored graph?' We answer this question with a mapping from classes and edge weights to a real number, and use Dijkstra's Algorithm to compute the best path. Correctness is proven and an implementation is highlighted.