Mark Witkowski

High-Level Robot Control through Logic

This paper presents a programmable logic-based agent control system that interleaves planning, plan execution and perception. In this system, a program is a collection of logical formulae describing the agents relationship to its environment. Two such programs for a mobile robot are described -- one for navigation and one for map building -- that share much of their code. The map building program incorporates a rudimentary approach to the formalisation of epistemic fluents, knowledge goals, and knowledge producing actions.

The inspection of very large images by eye-gaze control

The increasing availability and accuracy of eye gaze detection equipment has encouraged its use for both investigation and control. In this paper we present novel methods for navigating and inspecting extremely large images solely or primarily using eye gaze control. We investigate the relative advantages and comparative properties of four related methods: Stare-to-Zoom (STZ), in which control of the image position and resolution level is determined solely by the users gaze position on the screen; Head-to-Zoom (HTZ) and Dual-to-Zoom (DTZ), in which gaze control is augmented by head or mouse actions; and Mouse-to-Zoom (MTZ), using conventional mouse input as an experimental control. The need to inspect large images occurs in many disciplines, such as mapping, medicine, astronomy and surveillance. Here we consider the inspection of very large aerial images, of which Google Earth is both an example and the one employed in our study. We perform comparative search and navigation tasks with each of the methods described, and record user opinions using the Swedish User-Viewer Presence Questionnaire. We conclude that, while gaze methods are effective for image navigation, they, as yet, lag behind more conventional methods and interaction designers may well consider combining these techniques for greatest effect.

Advances in Autonomous Robotics

In order to perform a reaching movement towards a moving target, an autonomously learning robot must first learn several transformations, such as motion detection, coordinate transformation between the camera and the arm and the inverse model of the arm. A curious reaching robot learns them better by performing the appropriate actions so as to expedite and improve their learning speed and accuracy. We implement a model of hierarchical curiosity loops in an autonomous active learning robot, whereby each loop converges to the optimal action that maximizes the robot’s learning of the appropriate transformation. It results in the emergence of unique behaviors that ultimately lead to the capability of reaching.

Agent mediated retailing in the connected local community

Abstract This paper reports on and discusses recent research into software agent mediated retailing and considers how it may act as a catalyst to the formation of local electronic retail communities. Each member of the e-retail community has their own software agent that acts on their behalf. Their software agent acts as an electronic personal sales assistant (ePSA), interacting with the person and assisting with the selection and purchase of products. We consider issues that arise when each person can express their likes and dislikes for various aspects of the product range to their software agent, both in terms of the software agents ability to serve them better, and how such preferences may be shared with others in the larger community to help them—and others in the community—make better and more informed choices. The investigations reported use the multimedia access through personal persistent agents (MAPPA) system, an experimental kiosk based e-retail system. MAPPA integrates a software agent based architecture, a novel form of product display and the use of an animated character agent to enhance the users sense of personalisation. We report on user evaluation studies of the MAPPA system to investigate the effectiveness of the character agent as an interface to the e-retail system using both conventional usability evaluation techniques and eye-tracking technology. We also describe a simulation program that allows us to investigate the dynamics of e-retail community cohesion when a number of different parameters are considered, including the characteristics of the individual community members and the algorithms employed by the software agents to share information. Lastly, we develop and discuss notions of loyalty , trust , reputation and preference , four issues critical to the development and maintenance of the relationships between individual people, their personal software agents, businesses and the larger e-retail connected community. We propose these issues as ‘supra-functional’ criteria for human–computer interaction design in this area.

An Action-Selection Calculus

This article describes a unifying framework for five highly influential but disparate theories of natural learning and behavioral action selection. These theories are normally considered independently, with their own experimental procedures and results. The framework presented builds on a structure of connection types, propagation rules and learning rules, which are used in combination to integrate results from each theory into a whole. These connection types and rules form the action-selection calculus. The calculus will be used to discuss the areas of genuine difference between the factor theories and to identify areas where there is overlap and where apparently disparate findings have a common source. The discussion is illustrated with exemplar experimental procedures. The article focuses on predictive or anticipatory properties inherent in these action-selection and learning theories, and uses the dynamic expectancy model and its computer implementation SRS/E as a mechanism to conduct this discussion.

A comparison of image inspection modes for a visual search and rescue task

Visual inspection and search are important tasks in many fields, including quality control, security surveillance and medical diagnosis. We investigate whether it is better to visually inspect a moving image as opposed to a series of equivalent static images using the challenging problem of locating individuals lost in a wilderness. Wilderness search and rescue may be approached with a systematic aerial search assisted by an unmanned aerial vehicle (UAV) whose camera relays the terrain below for human inspection. We investigated two presentation modes of simulated UAV video feeds. The first mimics the live video from the downward facing camera. In the second, ‘serial visual presentation’ (SVP) mode, a static image remains in view until replaced by a new image at a rate equivalent to the live video mode. We established a statistically significant improvement in the number of detected targets in the SVP mode when compared to the Moving mode. However, these improvements were accompanied by an increase in the number of incorrectly identified targets in SVP mode. UAV speed has a significant effect on target identification in both modes, presumably due to the extra time available for viewing at lower speeds. We found no significant interaction between speed and presentation mode.

Energy Sharing for Swarms Modeled on the Common Vampire Bat

This article reports on observations arising from the use of a simulator in the conteXt of energy sharing (trophallaXis) that might be applied to swarms or communities of energy critical robots. Using food-sharing behavior reported for the common vampire bat species Desmodus rotundus, it is presumed that such robots would gather naturally occurring but limited fuel materials and convert them to energy as required. Energy sharing can bring substantial benefits to the group as a whole in situations where suitable materials are encountered unevenly by members of the population, in a manner insufficient to guarantee the continuing survival of any individual. Using the simulator the survivability of groups is investigated under various regimes, including sharing and non-sharing, the effects of energy dependents, group size and non-sharing defaulters. The role of reciprocation in feeding is also considered.

A dialectic architecture for computational autonomy

This paper takes the view that to be considered autonomous, a software agent must possess the means by which to manage its own motivations and so arbitrate between competing internal goals. Using the motivational theories of Abraham Maslow as a starting point, we investigate the role that argumentation processes might play in balancing the many competing aspects of a whole agents motivational agenda. This is developed into an Agent Argumentation Architecture (AAA) in which multiple faculties argue for different aspects of the total behavior of the Agent. The overall effect of these internal arguments then defines which actions the agent will select for expression, and so define the overt and observable personality of the agent.

Towards a Four Factor Theory of Anticipatory Learning

This paper takes an overtly anticipatory stance to the understanding of animat learning and behavior. It analyses four major animal learning theories and attempts to identify the anticipatory and predictive elements inherent to them, and to provide a new unifying approach based on the anticipatory nature of those elements based on five simple predictive ”rules”. These rules encapsulate all the principal properties of the four diverse theories (the four factors) and provide a simple framework for understanding how an individual animat may appear to operate according to different principles under varying circumstances. The paper then indicates how these anticipatory principles can be used to define a more detailed set of postulates for the Dynamic Expectancy Model of animat learning and behavior, and to construct its computer implementation SRS/E. Some of the issues discussed are illustrated with an example experimental procedure using SRS/E

Towards Autonomous Robotic Systems

Robot with non-back-drivable actuators will appear stiff when in contact with the environment and human. This scenario is unsafe for the Human-Robot Interaction (HRI). In order to guarantee safety in HRI, the robot will be made “soft” such that a compliant control can be introduced. Apart from utilizing the proper mechanism design, the back drivability actuators can be achieved by a suitable choice of control. In particular, in this paper, a PID control is employed to achieve an active compliance control. The reference impedance model characteristics are exploited for which the system allows us to introduce a virtual mass-spring-damper system to adjust the compliant control level. The performance of the PID control will be tested on the RED Hand in the simulation. The results are recorded and analyzed for the thumb finger. The results show that the PID controller is capable of controlling the motion and position of the RED Hand. In addition, the compliance behavior for the RED Hand can be suitably adjusted based on the required compliant level.