Papers by Mary-Anne Williams
Logical formalization of commonsense reasoning : papers from the 2003 AAAI Symposium, March 24-26, Stanford, California
AAAI Press eBooks, 2003
arXiv (Cornell University), May 29, 2016
Python based Robot Interactive Development Environment (PyRIDE) is a software that supports rapid... more Python based Robot Interactive Development Environment (PyRIDE) is a software that supports rapid interactive programming of robot skills and behaviours on PR2/ROS (Robot Operating System) platform. One of the key features of PyRIDE is its interactive remotely accessible Python console that allows its users to program robots online and in realtime in the same way as using the standard Python interactive interpreter. It allows programs to be modified while they are running. PyRIDE is also a software integration framework that abstracts and aggregates disparate low level ROS software modules, e.g. arm joint motor controllers, and exposes their functionalities through a unified Python programming interface. PR2 programmers are able to experiment and develop robot behaviours without dealing with specific details of accessing underlying softwares and hardwares. PyRIDE provides a client-server mechanism that allows remote user access of the robot functionalities, e.g. remote robot monitoring and control, access real-time robot camera image data etc. This enables multi-modal human robot interactions using different devices and user interfaces. All these features are seamlessly integrated into one lightweight and portable middleware package. In this paper, we use four real life scenarios to demonstrate PyRIDE key features and illustrate the usefulness of software.
Intelligent agent based architecture for internet marketing
Lecture Notes in Computer Science, 2001
On Designing Socially Acceptable Reward Shaping
Lecture Notes in Computer Science, 2016
For social robots, learning from an ordinary user should be socially appealing. Unfortunately, ma... more For social robots, learning from an ordinary user should be socially appealing. Unfortunately, machine learning demands an enormous amount of human data, and a prolonged interactive teaching session becomes anti-social. We have addressed this problem in the context of reward shaping for reinforcement learning. For efficient reward shaping, a continuous stream of rewards is expected from the teacher. We present a simple framework which seeks rewards for a small number of steps from each of a large number of human teachers. Therefore, it simplifies the job of an individual teacher. The framework was tested with online crowd workers on a transport puzzle. We thoroughly analyzed the quality of the learned policies and crowd’s teaching behavior. Our results showed that nearly perfect policies can be learned using this framework. The framework was generally acceptable in the crowd’s opinion.
Innate and Learned Emotion Network
Autonomous agents sometimes can only rely on the subjective information in terms of emotions to m... more Autonomous agents sometimes can only rely on the subjective information in terms of emotions to make decision due to the inavailability of the nonsubjective knowledge. However, current emotion models lack of integrating innate emotion and learned emotion and tend to focus on a specific aspect. This paper describes the underlying new computational emotion model in ASMO which integrates both innate and learned emotions as well as reasoning based on probabilistic causal network. ASMO’s emotion model is compared with other models and related works and shows its practical capabilities to utilize subjective knowledge in decision making.
Decision-Theoretic Human-Robot Interaction: Designing Reasonable and Rational Robot Behavior
Lecture Notes in Computer Science, 2016
Autonomous robots are moving out of research labs and factory cages into public spaces; people’s ... more Autonomous robots are moving out of research labs and factory cages into public spaces; people’s homes, workplaces, and lives. A key design challenge in this migration is how to build autonomous robots that people want to use and can safely collaborate with in undertaking complex tasks. In order for people to work closely and productively with robots, robots must behave in way that people can predict and anticipate. Robots chose their next action using the classical sense-think-act processing cycle. Robotists design actions and action choice mechanisms for robots. This design process determines robot behaviors, and how well people are able to interact with the robot. Crafting how a robot will choose its next action is critical in designing social robots for interaction and collaboration. This paper identifies reasonableness and rationality, two key concepts that are well known in Choice Theory, that can be used to guide the robot design process so that the resulting robot behaviors are easier for humans to predict, and as a result it is more enjoyable for humans to interact and collaborate. Designers can use the notions of reasonableness and rationality to design action selection mechanisms to achieve better robot designs for human-robot interaction. We show how Choice Theory can be used to prove that specific robot behaviors are reasonable and/or rational, thus providing a formal, useful and powerful design guide for developing robot behaviors that people find more intuitive, predictable and fun, resulting in more reliable and safe human-robot interaction and collaboration.
Smooth robot motion with an Optimal Redundancy Resolution for PR2 robot based on an analytic inverse kinematic solution
Current support for the PR2 humanoid arm motion planning does not provide a redundancy resolution... more Current support for the PR2 humanoid arm motion planning does not provide a redundancy resolution scheme and this can cause excessive movements and discontinuities in the generated trajectories. We provide an innovative solution and implementation in ROS that can be used to deliver smooth motions that people find to be intuitive and easy to predict. This paper, provides a comprehensive analysis of kinematics of the PR2 humanoid robot arm including an analytic Inverse Kinematics solution and an Optimal Redundancy Resolution scheme. First, a closed form IK computation method is introduced for the PR2 arms providing all feasible solutions in global jointspace for a given value of the first joint angle as a redundant parameter. Then, a redundancy optimization technique is customized and formulated based on a desired objective function that finds optimal values for the redundant parameter. The proposed technique computes robot motion plan more effectively so that the robot behaviors are expected to be more reliable. The technique has been implemented successfully on PR2 for a hand writing task.
Fusion et révision de bases de connaissances avec des priorités
HAL (Le Centre pour la Communication Scientifique Directe), 1999
Proceedings of the 2006 international symposium on Practical cognitive agents and robots
Pacific Asia Conference on Information Systems, 2011
Data purpose is a central concept to modeling privacy requirements for information systems. Exist... more Data purpose is a central concept to modeling privacy requirements for information systems. Existing purpose-centric approaches for privacy protection have mainly focused on access control. The problem of ensuring the consistency between data purpose and data usage has been under-addressed. Given the lack of practical purpose-centric solutions, we argue that a grounded understanding of the underlying concepts of data purpose and usage is fundamental to modeling privacy requirements. In recognition of an existing "privacy rights" framework, this paper develops an ontological grounding of data purpose and usage that can be used to understand their implications on fundamental privacy rights for modeling privacy requirements for information systems.
Explainable artificial intelligence
Edward Elgar Publishing eBooks, May 14, 2021
Principles of knowledge representation and reasoning : proceedings of the Ninth International Conference (KR2004), Whistler, BC, Canada, June 2-5, 2004
AAAI Press eBooks, 2004
Lecture Notes in Computer Science, 2016
A critical aspect of rational belief revision that has been neglected by the classical AGM framew... more A critical aspect of rational belief revision that has been neglected by the classical AGM framework is what we call the principle of kinetic consistency. Loosely speaking, this principle dictates that the revision policies employed by a rational agent at different belief sets, are not independent, but ought to be related in a certain way. We formalise kinetic consistency axiomatically and semantically, and we establish a representation result explicitly connecting the two. We then combine the postulates for kinetic consistency, with Parikh's postulate for relevant change, and add them to the classical AGM postulates for revision; we call this augmented set the extended AGM postulates. We prove the consistency and demonstrate the scope of the extended AGM postulates by showing that a whole new class of concrete revision operators introduced hererin, called PD operators, satisfies all extended AGM postulates. PD operators are of interest in their own right as they are natural generalisations of Dalal's revision operator. We conclude the paper with some examples illustrating the strength of the extended AGM postulates, even for iterated revision scenarios.
arXiv (Cornell University), Jul 29, 2015
Action recognition from still images is an important task of computer vision applications such as... more Action recognition from still images is an important task of computer vision applications such as image annotation, robotic navigation, video surveillance and several others. Existing approaches mainly rely on either bag-of-feature representations or articulated body-part models. However, the relationship between the action and the image segments is still substantially unexplored. For this reason, in this paper we propose to approach action recognition by leveraging an intermediate layer of "superpixels" whose latent classes can act as attributes of the action. In the proposed approach, the action class is predicted by a structural model(learnt by Latent Structural SVM) based on measurements from the image superpixels and their latent classes. Experimental results over the challenging Stanford 40 Actions dataset report a significant average accuracy of 74.06% for the positive class and 88.50% for the negative class, giving evidence to the performance of the proposed approach 1 .
Internet users and internet shopping: individual perceptions and consumer behavior
The fugitive
The aim of the movie is to highlight some of the key challenges facing so-cial robots in the wild... more The aim of the movie is to highlight some of the key challenges facing so-cial robots in the wild. The opening scene shows a PR2 leaving research laboratory venturing into the real world alone in search of meaning. Each subsequent scene in the movie raises important research questions highlighting problems that need to be addressed in the field of social service robotics. When will robots wander around buildings unsupervised? How will they navigate and localize with glass walls: this research problem is exposed when a robot finds itself having to move around a real building. The robot is independent and has a sense of self. It wants to engage in society. It solves this problem by finding a job in a cafe where it is as-signed menial tasks, but aspires to be a barista. Thus raising the question of whether PR2 robots are suited to working with hot steaming liquids. Still the robot can dream, why not. The robot realizes in order to progress it needs to learn some new skills and it is shown teaching itself a new skill and practicing to improve its performance. When it is time to put the new skill into practice, the robot has a revelation, discovering in the act of doing that there can be preconditions attached to the enaction of skills, i.e. people do not need peanut butter until they have bread to spread it on. The robot demonstrates his robust understanding of social etiquette by not only offering the peanut butter to the female-human first, but also chastising a male-human for not observing this important social protocol. The story ends with the recaptured robot being dragged back to the lab. The robot appears to be mortified by its loss of freedom and looks utterly dejected and dispirited. The robot’s behavior generates empathy the human minder, but the robot is pretending to be disheartened, and is deceit-fully planning its next escapade as a Jedi Knight! Deception is a highly sophisticated cognitive skill: a capability enabled by a theory of mind which is necessary for communication, social interaction and collaboration, all critically important skills for a service robot. Categories and Subject Descriptors J.4 [Social and Behavioral Sciences]: Sociology, Psychology
A Distributed Future:How Blockchain Affects Strategic Management, Organisation Design & Governance
Proceedings - Academy of Management, Aug 1, 2017
Blockchain is a new technology that transforms strategic management, organisational design and go... more Blockchain is a new technology that transforms strategic management, organisational design and governance due to its decentralised and distributed characteristics. It is a database technology, a di...
Journal of Artificial Intelligence Research, Nov 26, 2019
In this article, the epistemic-entrenchment and partial-meet characterizations of Parikh's releva... more In this article, the epistemic-entrenchment and partial-meet characterizations of Parikh's relevance-sensitive axiom for belief revision, known as axiom (P), are provided. In short, axiom (P) states that, if a belief set K can be divided into two disjoint compartments, and the new information ϕ relates only to the first compartment, then the revision of K by ϕ should not affect the second compartment. Accordingly, we identify the subclass of epistemic-entrenchment and that of selection-function preorders, inducing AGM revision functions that satisfy axiom (P). Hence, together with the faithful-preorders characterization of (P) that has already been provided, Parikh's axiom is fully characterized in terms of all popular constructive models of Belief Revision. Since the notions of relevance and local change are inherent in almost all intellectual activity, the completion of the constructive view of (P) has a significant impact on many theoretical, as well as applied, domains of Artificial Intelligence. 1. To distinguish the research area from the process, we shall use the capitalized term "Belief Revision" for the former, and the same term in lower-case letters (i.e., "belief revision") for the latter.
Uploads
Papers by Mary-Anne Williams