embodied agents
Teaching Embodied Agents, Applied to Virtual Animals in Second Life
Ben Goertzel presents at AGI-08 on a paper by the speaker, Cassio Pennachin, Nil Geissweiller, Moshe Looks, Andre Senna, Welter Silva, Ari Heljakka, and Carlos Lopes. A teaching methodology called Imitative-Reinforcement-Corrective (IRC) learning is described, and proposed as a general approach for teaching embodied non-linguistic AGI systems. IRC may be used with a variety of different learning algorithms, but it is particularly easily described in EC lingo. In these terms, it is a framework for automatically learning a procedure that generates a desired type of behavior, in which: a set of exemplars of the target behavior-type are utilized for fitness estimation; reinforcement signals from a human teacher are used for fitness evaluation; and the execution of candidate procedures may be modified by the teacher via corrections delivered in real-time. AGI-08 camera, editing: Jeriaska
Embodied Conversational Interface Agent at Chatbots.org | Project RED
Louise, a part of the Project RED, is one of the Virtual Patient Advocates which assists the Discharge Advocates in teaching patients about components of their care, such as their prescribed medications, follow-up appointments and diagnoses. Read more: www.chatbots.org
Seminário LabMAg: odNEAT - Algorithm for Distributed Online, Onboard Evolution of Robot Behaviours
Seminário LabMAg odNEAT: An Algorithm for Distributed Online, Onboard Evolution of Robot Behaviours Abstract: We propose and evaluate a novel approach called Online Distributed NeuroEvolution of Augmenting Topologies (odNEAT). odNEAT is a completely distributed evolutionary algorithm for collective online learning in groups and swarms of embodied agents such as robots. Each agent is controlled by an artificial neural network (ANN) that represents a candidate solution to a given task. The candidate solutions undergo a decentralized evolution of both connection weights and network topology. The embodied agents communicate their solutions to nearby neighbours from time to time. We demonstrate odNEAT through a series of simulation-based experiments in which a group of e-puck-like robots must perform an aggregation task. Our results show that the robots are capable of evolving effective search strategies and sustainable behaviours evolve quickly. We analyse the contribution of each algorithmic component on performance through a series of ablation studies. Fernando Silva, LabMAg, Faculdade de Ciências da Universidade de Lisboa
WallFollower attempt 1
this is a video of our first attempt to make a Wall follower as part of lab session 2 in the course Embedded Systems - Embodied Agents, Digital Control in an Physical World at Daimi.
A Cognitive Model for Embodied Gesture Processing in Virtual Agents
Human social interaction encompasses verbal and nonverbal aspects in which hand gesture is a widely used nonverbal communication way. Handling gestures requires different cognitive and knowledge levels, from motor skills to social intentions behind those movements. During social interaction all of these levels can be used for cognitive processes of both perception and generation, as evidenced by neuropsychological studies. In this video, we focus on the motor skills of gestures at different levels of abstraction, from movement trajectories towards higher levels of semantics. In this context, we aim to engage humanoid virtual agents in social interaction with humans. To this end, we developed a cognitive computational model which tries to capture the embodied basis of hand gestures and we demonstrate how it enables to combine and bootstrap the online learning, perception, recognition and generation of gestural movements in a human-agent interaction scenario. Research website: www.techfak.uni-bielefeld.de
Expressive Conversational Agent in Augmented Reality
Virtual characters are an emerging interface metaphor among the human computer interaction research community. The use of a virtual character as an interface simulates, as possible, the seamless natural interaction between the user and the computer. In this video we demonstrate our work of incorporating an expressive conversational virtual agent (Greta) into an Augmented Reality (AR) environment to allow it to inhabit the real world with the user. To incorporate the agent into an AR environment, we use the ARToolKit computer vision library. The AR-Player permits to use more than one character in the scene. Each of them can become an independent agent with its own behavior specification, and its own instantiation of Behavior Planner. These different agents can be associated with different markers in the same scene. The AR character can be scaled to create different impressions, like a small character sitting on the desk in front of you, or a life-size character to simulate real world interactions.
Embodied Conversational Agents with Environment-, Self- and Interaction-Awareness
Research on embodied conversational agents' reasoning and actions has mostly ignored the external environment. We argue that believability of such agents is tightly connected with their ability to relate to the environment during a conversation. This ability, defined as awareness believability, is formalised in terms of three components - environment-, self- and interaction-awareness. The demo presents a method enabling virtual agents to reason about their environment, understand the interaction capabilities of other participants, own goals and current state of the environment, as well as to include these elements in conversations.
The LEGO Turing Machine
A TV Shop themed demonstration of a Turing Machine made in LEGO Mindstorms. It was made as part of a project at computer science at Aarhus University. A blog about the project is available at legoofdoom.blogspot.com
GALA demo - A BML-based Embodied Conversational Agent for a Personality Detection Program
The psychometer is a validated instrument for the measurements of personality traits of a human being. To determine the personality of the user, the psychometer asks a set of questions using natural language. It allows users to ask for clarifications of the question and asks follow-up questions if the answer from the user cannot readily be classified in one of the 5 categories it needs. The latest version of the psychometer is implemented as a talking head using Elckerlyc by Guillermo Solano.
VicTour, Virtual Interactive Character Tour Guide
VicTour is YLabs 's final proof-of-concept resulting from its latest software research project CHAMELEON. During the development process the project generated other examples already published such as FlyAR ( www.youtube.com and the "mother of all depth sensing demos", long before the whole Kinect frenzy (www.youtube.com This depth sensing demo was distinguished with the first Auggie award (the Augmented Reality Oscars) at the Augmented Reality Event 2010. CHAMELEON emerged from the strong belief that the research and development of next-generation intelligent interaction devices is expected to rely on integrative efforts across several research fields. The main objective of this project was therefore to explore and implement architectures and practical design methodologies for embodied intelligent interaction, with a focus on affective and cognitive computation models, as well as autonomous adaptation and learning of system components and parameters over dynamic multi-modal and multi-user environments. More specifically, research focused on: • Situated cognition and human action models to support development of natural interaction systems; • Embodied agent architectures suitable to the scalable integration of learning and affective computing algorithms; • Unsupervised, reinforcement and evolutionary learning techniques for autonomous agents in the context of multi-modal natural interaction installations; • Affective computing techniques for the design of emotion-based ...
ECABank a web based moch banking system with an Embodied Conversational Agent (ECA)
Serious Games research by D. Romano at University of Sheffield, UK. Research porject by Orn Gurnasson and Daniela Romano see www.dcs.shef.ac.uk
LEGO NXT using ZigBee 1 (Robot view)
The goal is to make the robots stick together in a flock using the signal strength of the radio signals. The NXT runs LEJOS and has a ZigBee board connected through I2C. A very simple triangulation algortihm is used. Final project in the course "Embedded Systems - Embodied Agents, Digital Control in an Physical World" at Aarhus University 2009. DevCom ZigBee modules using Freescale chipsets are provided by Develco A/S. The video is recorded with a robot mounted HTC Tattoo.
LEGO NXT using ZigBee 2 (Robot view)
The goal is to make the robots stick together in a flock using the signal strength of the radio signals. The NXT runs LEJOS and has a ZigBee board connected through I2C. A very simple triangulation algortihm is used. Final project in the course "Embedded Systems - Embodied Agents, Digital Control in an Physical World" at Aarhus University 2009. DevCom ZigBee modules using Freescale chipsets are provided by Develco A/S. The video is recorded with a robot mounted HTC Tattoo.
Balancing robot 1
An attempt at making a robot that can balance on two wheels. Not particularly successful, but not a complete failure either, it can keep its balance for some 5 seconds before overbalancing. Build as part of the Embedded Systems - Embodied Agents course at Aarhus University.
LEGO NXT using ZigBee 3
The goal is to make the robots stick together in a flock using the signal strength of the radio signals. The NXT runs LEJOS and has a ZigBee board connected through I2C. A very simple triangulation algortihm is used. Final project in the course "Embedded Systems - Embodied Agents, Digital Control in an Physical World" at Aarhus University 2009. DevCom ZigBee modules using Freescale chipsets are provided by Develco A/S. The video is recorded with a HTC Hero.
Balancing robot 2
An attempt at making a robot that can balance on two wheels. Not particularly successful, but not a complete failure either, it can keep its balance for some 5 seconds before overbalancing. Build as part of the Embedded Systems - Embodied Agents course at Aarhus University.
Binh Nguyen - Evolved AI, embodiment in virtual worlds - Singularity Summit Australia 2010 09
"Evolving Artificial Intelligences with Human Like Embodiments and Environments" There are a number of breathtaking projects underway including the Blue Brain Project led by Dr. Henry Markram, the China Brain Project formerly led by Dr. Hugo De Garis, and the OpenCog Project led by Dr. Ben Goertzel. They are breathtaking in their vision and complexity. This talk is instead about a relatively simple idea. What if instead of scanning and modelling biological neural networks with ever more detail, or building neural networks with a staggering number of neurons, or designing cognitive architectures with immense knowledge databases, we did something different? What if we instead evolve artificial intelligences by requiring them to learn to survive with increasingly human like bodies and environments? Skilfully coping with the world requires a lifetime of experiences. It requires engaging in a form of life, interacting with the physical world and with others in the social world. Artificial intelligences are evolving behaviours that are ever more sophisticated. Behaviours include walking, hunting, evading, foraging, reproducing, cooperating, competing, using objects, and altering environments. Most interestingly, artificial intelligences are evolving natural language, one of the most powerful human abilities. The talk will touch upon a few key ideas. Firstly, life is an incredibly persistent force given the chance, and persistence itself can compel artificial intelligences to ...
Trenutak spoznaje: Embodied Conversational Agents
Embodied Conversational Agents (Croatian TV, 28.05.2004.) In the interview for the science program "Trenutak spoznaje" of the Croatian Television, we discuss the Embodied Conversational Agents (ECAs), their implementation and applications. We demonstrate our current implementations at the time of interview.
Marve: A Prototype Virtual Human Interface Framework for Studying Human-Virtual Human Interaction
Human to virtual human interaction is the next frontier in interface design, particularly for tasks that are social or collaborative in nature. Several embodied interface agents have been developed for specific social, place-related tasks, but empirical evaluations of these systems have been rare. In this work, we present Marve (Messaging And Recognition Virtual Entity), our general purpose Virtual Human Interface Framework, which integrates cutting-edge interface technologies into a seamless real-time system, to study human to virtual human interaction. Marve is a prototype of a real-time embodied, interactive, autonomous, virtual human interface agent framework. Marve "lives" next to the primary entrance of the Future Computing Lab. His primary tasks are to greet everyone who enters or leaves the lab, and to take and deliver messages to the students and faculty who work there. Marve uses computer vision techniques for passer-by detection, gaze tracking, and face recognition, and communicates via natural language. We present a preliminary empirical study of the basic elements of Marve, including interaction response times, recognition of friends, and ability to learn to recognize new people.
Balancing robot 3
An attempt at making a robot that can balance on two wheels. Not particularly successful, but not a complete failure either, it can keep its balance for some 5 seconds before overbalancing. Build as part of the Embedded Systems - Embodied Agents course at Aarhus University.
From Bricks to Brains: What is embodied cognitive science?
From Bricks to Brains introduces embodied cognitive science, and illustrates its foundational ideas through the construction and observation of LEGO Mindstorms robots. Discussing the characteristics that distinguish embodied cognitive science from classical cognitive science, From Bricks to Brains places a renewed emphasis on sensing and acting, the importance of embodiment, the exploration of distributed notions of control, and the development of theories by synthesizing simple systems and exploring their behaviour. Numerous examples are used to illustrate a key theme: the importance of an agents environment. Even simple agents, such as LEGO robots, are capable of exhibiting complex behaviour when they can sense and affect the world around them. Authors: Michael Dawson , Brian Dupuis and Michael Wilson (University of Alberta)
LEGO NXT using ZigBee 4
The goal is to make the robots stick together in a flock using the signal strength of the radio signals. The NXT runs LEJOS and has a ZigBee board connected through I2C. A very simple triangulation algortihm is used. Final project in the course "Embedded Systems - Embodied Agents, Digital Control in an Physical World" at Aarhus University 2009. DevCom ZigBee modules using Freescale chipsets are provided by Develco A/S. The video is recorded with a HTC Hero.
Balancing robot 4
Another attempt at making a robot that can balance on two wheels. This is an early test run where the robot managed to stand upright for a minute or so. Build as part of the Embedded Systems - Embodied Agents course at Aarhus University.
LEGO NXT using ZigBee 5
The goal is to make the robots stick together in a flock using the signal strength of the radio signals. The NXT runs LEJOS and has a ZigBee board connected through I2C. A very simple triangulation algortihm is used. Final project in the course "Embedded Systems - Embodied Agents, Digital Control in an Physical World" at Aarhus University 2009. DevCom ZigBee modules using Freescale chipsets are provided by Develco A/S. The video is recorded with a HTC Hero.
Balancing robot 5
Another attempt at making a robot that can balance on two wheels. This is the final test run where the robot managed to stand upright for more than four minutes. Build as part of the Embedded Systems - Embodied Agents course at Aarhus University.
Balancing robot 6
Another attempt at making a robot that can balance on two wheels. This is a video that show how the robot-computer interface works. The graph that's displayed on the whiteboard in the background is live data from the robot. Those of you who can understand Danish can also gain an insight into how we modified the behavior of the robot during the test runs. Yes, I know - the video is a bit dark. The reflection from the whiteboard played havoc with my camera... Build as part of the Embedded Systems - Embodied Agents course at Aarhus University.
Balancing robot 7
Another attempt at making a robot that can balance on two wheels. This video can give those of you who can understand Danish an insight into how we modified the behavior of the robot during the test runs. Build as part of the Embedded Systems - Embodied Agents course at Aarhus University.
Animated agents in a UbiComp environment
Most of today's Augmented Reality (AR) systems operate as passive information browsers relying on a finite and deterministic world model and a predefined hardware and software infrastructure. We propose an AR framework that dynamically and proactively exploits hitherto unknown applications and hardware devices, and adapts the appearance of the user interface to persistently stored and accumulated user preferences. Our framework explores proactive computing, multi-user interface adaptation, and user interface migration. We employ mobile and autonomous agents embodied by real and virtual objects as an interface and interaction metaphor, where agent bodies are able to opportunistically migrate between multiple AR applications and computing platforms to best match the needs of the current application context. We present two pilot applications to illustrate design concepts
Linefollower
A robot that follows a line. Sounds easy... Until you have to do it at speed. This robot completed the course in 21 seconds, 7 seconds slower than the current record. Build as part of the Embedded Systems - Embodied Agents course at Aarhus University.
Embodied Simulation of the Cyclic Pursuit Algorithm: Convergence of UAVs to a Stack Formation
Cyclic pursuit is a decentralized control algorithm used to achieve collaborative operations in multi-agent systems. It mimics the behaviour of biological organisms such as birds, beetles, etc. Every agent pursues the agent next to it with a speed proportional to the distance between the two agents and the velocity vector pointing towards the agent. Implementation of these individual rules leads to the convergence of agents to the centroid of their initial positions. The agents are modeled as UAVs in the Player/Gazebo environment (3D Simulator for embodied robotic simulations). Convergence to various formations under these constraints is analyzed. For more information: www.ee.ucr.edu
Bugs
Bugs are general purpose robotic platforms designed for experimentation with ideas of swarm intelligence. They are physically embodied autonomous agents interacting as an artificial life system. Each robot is programmed with simple rules such as "seek light" and "avoid obstacles". Through their interaction with each other and the environment these simple low level behaviors give rise to interesting higher level patterns of organization. This video is a studio demo. It shows a test of 5 agents exploring flocking behavior at 3x speed.
Voice-enabled Jukebox application (shown at FET 2009)
Interactive demo of jukebox (mp3 player) application controlled by free-form voice commands. The user interface is enriched by embodied conversational agent in the form of 3D model of human head where lips movement is synchronized with text-to-speech generation.
A multiuser ECA system
The video shows the results of the ongoing work in which mixed Japanese-Croatian research team investigates principles of integrating a multiuser support in an Embodied Conversational Agents (ECA) system. It is a prototype system in which a tour guide ECA interacts with one or two users. The system combines different technologies to detect and address the system users and draw their attention. Experimental interaction with the system produces encouraging results. The system can address the user's appearance, departure, decreased level of interest and identify his conversational role. The project has been done in collaboration between University of Kyoto, University of Zagreb and University of Tokyo.
White House Briefing on Federal Agency Operational Plans to Implement the National HIV/AIDS Strategy
Update on the work of the Federal agencies since the release of the National HIV/AIDS Strategy and to discussion of the President's HIV/AIDS priorities as embodied in his fiscal year 2012 Budget.
MARVE demo
Human to virtual human interaction is the next frontier in interface design, particularly for tasks that are social or collaborative in nature. Several embodied interface agents have been developed for specific social, place-related tasks, but empirical evaluations of these systems have been rare. In this work, we present Marve (Messaging And Recognition Virtual Entity), our general purpose Virtual Human Interface Framework, which integrates cutting-edge interface technologies into a seamless real-time system, to study human to virtual human interaction. Marve is a prototype of a real-time embodied, interactive, autonomous, virtual human interface agent framework. Marve "lives" next to the primary entrance of the Future Computing Lab. His primary tasks are to greet everyone who enters or leaves the lab, and to take and deliver messages to the students and faculty who work there. Marve uses computer vision techniques for passer-by detection, gaze tracking, and face recognition, and communicates via natural language. We present a preliminary empirical study of the basic elements of Marve, including interaction response times, recognition of friends, and ability to learn to recognize new people.
Demo of Embodied Anaphor Resolution
This video demonstrates how embodied experience is used to enable an AI to correctly resolve references in natural language (in this case, to tell what object an instance of the word "it" refers to). The video was screen-captured from actual real-time interaction between a human and an AI in the Multiverse virtual world. The virtual dog is controlled by the AI; the human avatar is controlled by a human player. While a relatively simple example, it involves integration of a natural language engine (RelEx), a cognition engine (the OpenCog-based OpenPetBrain), and a virtual world engine (Multiverse). The text box that sometimes shows at the bottom left of the video contains the text typed by the human player, to communicate with the AI dog. The thought bubbles and captions shown elsewhere in the video, were added to the video after it was recorded, to better explain what is going on. For more information, please go to: www.opencog.org
ESEA Lesson 6
An NXT robot made in the Embedded Systems - Embodied Agents course at Århus University.
Unity3D sandbox prototype
A demo of a 3d sandbox that will eventually be used for embodied learning using OpenCog to control agents. Apologies for the camera view - the computer running the demo was too slow to record the video at the same time, so we had to resort to using an iPhone to record instead. Better videos will be in the pipeline (with OpenCog controlling the agents).
From Bricks to Brains: Feedback from the Worm
From Bricks to Brains introduces embodied cognitive science, and illustrates its foundational ideas through the construction and observation of LEGO Mindstorms robots. Discussing the characteristics that distinguish embodied cognitive science from classical cognitive science, From Bricks to Brains places a renewed emphasis on sensing and acting, the importance of embodiment, the exploration of distributed notions of control, and the development of theories by synthesizing simple systems and exploring their behaviour. Numerous examples are used to illustrate a key theme: the importance of an agents environment. Even simple agents, such as LEGO robots, are capable of exhibiting complex behaviour when they can sense and affect the world around them. Authors: Michael Dawson , Brian Dupuis and Michael Wilson (University of Alberta)
Innovation at SAP - Avatars at Work
www.sapphirenow.com This demonstration highlights exploratory work by SAP Labs in the area of embodied conversational agents, or avatars, in the workplace. Visit http for more on-demand, broadcast session replays from SAPPHIRE NOW.
From Bricks to Brains: Synthetic Psychology and the Element of Surprise
From Bricks to Brains introduces embodied cognitive science, and illustrates its foundational ideas through the construction and observation of LEGO Mindstorms robots. Discussing the characteristics that distinguish embodied cognitive science from classical cognitive science, From Bricks to Brains places a renewed emphasis on sensing and acting, the importance of embodiment, the exploration of distributed notions of control, and the development of theories by synthesizing simple systems and exploring their behaviour. Numerous examples are used to illustrate a key theme: the importance of an agents environment. Even simple agents, such as LEGO robots, are capable of exhibiting complex behaviour when they can sense and affect the world around them. Authors: Michael Dawson , Brian Dupuis and Michael Wilson (University of Alberta)
ESEA Lesson 7 (Vehicle 1)
Embedded systems, Embodied agents course at Aarhus University, Denmark. The robot moves forward, proportional to the colume it's sensor captures. The louder the noise, the faster it moves.
From Bricks to Brains: Robots Call the Shots
From Bricks to Brains introduces embodied cognitive science, and illustrates its foundational ideas through the construction and observation of LEGO Mindstorms robots. Discussing the characteristics that distinguish embodied cognitive science from classical cognitive science, From Bricks to Brains places a renewed emphasis on sensing and acting, the importance of embodiment, the exploration of distributed notions of control, and the development of theories by synthesizing simple systems and exploring their behaviour. Numerous examples are used to illustrate a key theme: the importance of an agents environment. Even simple agents, such as LEGO robots, are capable of exhibiting complex behaviour when they can sense and affect the world around them. Authors: Michael Dawson , Brian Dupuis and Michael Wilson (University of Alberta)
eHrvatska - Virtualni govorni agenti
Prof. Igor S. Pandzic and Aleksandra Cerekovic are discussing embodied conversational agents and present the ongoing work in HOTLab.
From Bricks to Brains: Why did you write this book and who will want to read it?
From Bricks to Brains introduces embodied cognitive science, and illustrates its foundational ideas through the construction and observation of LEGO Mindstorms robots. Discussing the characteristics that distinguish embodied cognitive science from classical cognitive science, From Bricks to Brains places a renewed emphasis on sensing and acting, the importance of embodiment, the exploration of distributed notions of control, and the development of theories by synthesizing simple systems and exploring their behaviour. Numerous examples are used to illustrate a key theme: the importance of an agents environment. Even simple agents, such as LEGO robots, are capable of exhibiting complex behaviour when they can sense and affect the world around them. Authors: Michael Dawson , Brian Dupuis and Michael Wilson (University of Alberta)
From Bricks to Brains: These Legs were made for Walking
From Bricks to Brains introduces embodied cognitive science, and illustrates its foundational ideas through the construction and observation of LEGO Mindstorms robots. Discussing the characteristics that distinguish embodied cognitive science from classical cognitive science, From Bricks to Brains places a renewed emphasis on sensing and acting, the importance of embodiment, the exploration of distributed notions of control, and the development of theories by synthesizing simple systems and exploring their behaviour. Numerous examples are used to illustrate a key theme: the importance of an agents environment. Even simple agents, such as LEGO robots, are capable of exhibiting complex behaviour when they can sense and affect the world around them. Authors: Michael Dawson , Brian Dupuis and Michael Wilson (University of Alberta)
From Bricks to Brains: Trailer
From Bricks to Brains introduces embodied cognitive science, and illustrates its foundational ideas through the construction and observation of LEGO Mindstorms robots. Discussing the characteristics that distinguish embodied cognitive science from classical cognitive science, From Bricks to Brains places a renewed emphasis on sensing and acting, the importance of embodiment, the exploration of distributed notions of control, and the development of theories by synthesizing simple systems and exploring their behaviour. Numerous examples are used to illustrate a key theme: the importance of an agents environment. Even simple agents, such as LEGO robots, are capable of exhibiting complex behaviour when they can sense and affect the world around them. Authors: Michael Dawson , Brian Dupuis and Michael Wilson (University of Alberta)
Page: 1 of 4
