Monday, July 31, 2006

Lab meeitng 3 August, 2006 (Chihao) : 2 Stereo-Microphone Source Location Methods

1. Stereo Microphone Sound Source Localization for Service Robots
Author: Axel Walthelm, Marek Litza
International Workshop on Advances in Service Robotics (ASER `03)
Abstract:
Service robots should have the ability of localizing acoustic sound sources especially when they have to work with and must be addressed by humans. A successful implementation of such an ability for the experimental mobile service robot MAVERIC is presented which is based on multi-scale cross-correlation of a stereo microphone signal.
Link

2. DUAL-MICROPHONE SOURCE LOCATION METHOD IN 2-D SPACE
Author: Weiwei Cui, Zhigang Cao, Jianqiang Wei
ICASSP 2006
Abstract:
Array based source location attracts a growing interest nowadays, which is frequently used in videoconference, hearing aids and hands free telephone systems to detect a speaker’s position. Time delay estimation (TDE) based dual-step source location is assumed to be classical method in this field. Recently, a new technique is developed based on interaural level difference (ILD) method, which determines the source position by the energy ratio from microphone pairs. However, all these localization techniques need at least three sensors to obtain a 2D source locus. In this work, we combine the former two techniques, i.e., ILD and TDE based techniques, to present a novel localization approach by an array of only two microphones, and further provide its closed form solution. Our final simulation confirms that such method, which is thought to be more suitable for the equipments with small size, can achieve a good result under normal conditions.
Link

Lab meeitng 3 August, 2006 (Zhen-Yu) : Educating C Language Using LEGO Mindstorms Robotic Invention System 2.0

ICRA2006

Author :
Seung Han Kim and Jae Wook Jeon

Abstract : A robot highly motivates students and it is one of
the best ways to connect students with the technology. LEGO
created a set called Robotic Invention System. This system helps
students to understand the technology of both robot and
programming language. It also improves creativeness by building
and controlling the robot. This paper will propose the idea on
educating C language to students using Robotic Invention System
2.0.
Index Terms - ANSI C Language, LEGO Mindstorms, Robotic
Command Explorer (RCX), Robotics Invention System (RIS)

link

Lab meeitng 3 August, 2006 (Stanley): Directed Sonar Sensing for Mobile Robot Navigation

Author: John J. Leonard and Hugh F. Durrant-Whyte

Abstract: This work investigates the problem of mobile robot navigation using sonar. We view model-based navigation as a process of tracking naturally occurring environment features, which we refer to as "targets". Targets that have been predicted from the environment map are tracked to provide vehicle position estimates. Targets that are observed, but not predicted, represent unknown environment features or obstacles, and cause new tracks to be initiated, classified, and ultimately integrated into the map.

link

News: The Wisdom of Robots

By David Cohn
10:00 AM Jul, 20, 2006

BOSTON -- You have to watch where you're walking at the artificial intelligence conference here this week -- you might trip over a roaming robot or bump into one flying around the room.

Celebrating the 50th anniversary of AI, the American Association for Artificial Intelligence has gathered leaders in the field from around the world to present research and give students a chance to square off in robot competitions.

See the full article.

Monday, July 24, 2006

Lab meeitng 27 July, 2006 (Eric): Real-Time Range Acquisition by Adaptive Structured Light

Thomas P. Koninckx and Luc Van Gool

Abstract—The goal of this paper is to provide a “self-adaptive” system for real-time range acquisition. Reconstructions are based on asingle frame structured light illumination. Instead of using generic, static coding that is supposed to work under all circumstances, systemadaptation is proposed. This occurs on-the-fly and renders the system more robust against instant scene variability and creates suitablepatterns at startup. A continuous trade-off between speed and quality is made. A weighted combination of different coding cues—basedupon pattern color, geometry, and tracking—yields a robust way to solve the correspondence problem. The individual coding cues areautomatically adapted within a considered family of patterns. The weights to combine them are based on the average consistency withthe result within a small time-window. The integration itself is done by reformulating the problem as a graph cut. Also, the cameraprojectorconfiguration is taken into account for generating the projection patterns. The correctness of the range maps is not guaranteed,but an estimation of the uncertainty is provided for each part of the reconstruction. Our prototype is implemented using unmodifiedconsumer hardware only and, therefore, is cheap. Frame rates vary between 10 and 25 fps, dependent on scene complexity.

[Link]

Lab meeitng 27 July, 2006 (Nelson): Multi-scale Point and Line Range Data Algorithms for Mapping and Localization

Multi-scale Point and Line Range Data Algorithms
for Mapping and Localization


Samuel T. Pfister and Joel W. Burdick

Division of Engineering and Applied Science
California Institute of Technology
Pasadena, California 91125, USA
Email:{sam,jwb}@robotics.caltech.edu

Abstract:
This paper presents a multi-scale point and line
based representation of two-dimensional range scan data. The
techniques are based on a multi-scale Hough transform and
a tree representation of the environment’s features. The multiscale
representation can lead to improved robustness and computational
efficiencies in basic operations, such as matching and
correspondence, that commonly arise in many localization and
mapping procedures. For multi-scale matching and correspondence
we introduce a χ2 criterion that is calculated from the
estimated variance in position of each detected line segment or
point. This improved correspondence method can be used as the
basis for simple scan-matching displacement estimation, as a part
of a SLAM implementation, or as the basis for solutions to the
kidnapped robot problem. Experimental results (using a Sick
LMS-200 range scanner) show the effectiveness of our methods.

Sunday, July 23, 2006

Stanford talk: Probability Collectives and Supervised Learning

David H. Wolpert
May 22, 2006, 4:15PM
http://graphics.stanford.edu/ba-colloquium/

Abstract
There are two major fields that analyze distributed systems: statistical physics and game theory. These fields can be re-expressed in a way that makes them mathematically identical. Doing so allows us to combine techniques from them, producing a hybrid formalism. That hybrid is called Probability Collectives (PC).

As borne out by numerous experiments, PC is particularly well-suited to black-box optimization and associated problems in distributed control. The core idea is that rather than directly optimize a variable of interest x, often it is preferable to optimize an associated probability distribution, P(x). That optimization can be done either via Monte Carlo Optimization (MCO) or, under certain circumstances, in closed form.

Recently it was realized that one can map MCO into a supervised machine learning problem. This means that all the powerful techniques of supervised learning can be used to improve MCO. As a special case, those techniques can be used to improve the optimization of P(x) in a PC-based optimizer. In this way the techniques of supervised learning can be leveraged to improve black-box optimization and distributed control.

In this talk I review PC. I also illustrate the identity between MCO and supervised learning using PC. In particular, I present results showing how cross-validation can be used to adaptively set an annealing schedule for the optimization of P(x), and to adaptively modify the complexity of P(x). I also illustrate the benefit of bagging in PC.

About the Speaker
David Wolpert received degrees in physics from the University of California, Santa Barbara, and from Princeton University. He is currently a Senior Computer Scientist at NASA and a consulting professor at Stanford. He was formerly head of a data mining group at IBM Almaden Research and a Postdoc at the Santa Fe Institute.

His work primarilly concerns how to design collectives of complex adaptive agents to achieve a global goal (i.e., distributed control and/or optimization). Other work investigates the bounds on computation that hold in broad classes of physical universes, and the use of self-dissimilarity as a complexity measure. He also works on extending game theory to sub-rational games using techniques from statistical physics.

MIT CSAIL Defense: Exploiting Biological Pathways to Infer Temporal Gene Interaction Models

Speaker: Corey Kemper , MIT CSAIL
Date: Friday, July 28 2006
Host: W.E.L. Grimson, John Fisher, MIT CSAIL

An important goal in genomic research is the reconstruction of the complete picture of temporal interactions among all genes, but this inference problem is not tractable because of the large number of genes, the small number of experimental observations for each gene, and the complexity of biological networks. We focus instead on the B cell receptor signaling pathway, which narrows the inference problem and provides a clinical application, as B cell chronic lymphocytic leukemia (B-CLL) is believed to be related to BCR response. In this work, we infer population-dependent gene networks of temporal interaction within the BCR signaling pathway. We develop simple statistical models that capture the temporal behavior of differentially expressed genes and then estimate the parameters in an Expectation-Maximization framework, resulting in clusters with a biological interpretation for each subject population. Using the cluster labels to define a small number of modes of interaction and imposing sparsity constraints to effectively limit the number of genes influencing each target gene makes the ill-posed problem of network inference tractable. For both the clustering and the inference of the predictive models, we have statistical results that show that we successfully capture the temporal structure of and the interactions between the genes relevant to the BCR signaling pathway. We have confirmatory results from a biological standpoint, in which genes that we have identified as playing key roles in the networks have already been shown in previous work to be relevant to BCR stimulation, but we also have results that guide future experiments in the study of other related genes, in order to further the long term goal of a full understanding of how and why B-CLL cells behave abnormally.

MIT CSAIL Defense: Coping with Uncertain Dynamics in Visual Tracking: Redundant State Models and Discrete Search Methods

Speaker: Leonid Taycher , CSAIL
Date: Thursday, July 27 2006
Host: Trevor Darrell, CSAIL

Abstract:
Model of the world dynamics is a vital part of any tracking algorithm. The observed world can exhibit multiple complex dynamics at different spatio-temporal scales. Faithfully modeling all motion constraints in a computationally efficient manner may be too complicated or completely impossible. Resorting to use of approximate motion models complicates tracking by making it less robust to unmodeled noise and increasing running times.

We propose two complimentary approaches to tracking with approximate dynamic models in probabilistic setting. The Redundant State Multi-Chain Model formalism described in the first part of the thesis allows combining multiple weak motion models, each representing a particular aspect of overall dynamic in a cooperative manner to improve state estimates. This is applicable, in particular, to hierarchical machine vision systems that combine trackers at several spatio-temporal scales. In the second part of the dissertation we propose supplementing exploration of the continuous likelihood surface by the discrete search in a fixed set of points distributed through the state space.

We demonstrate the utility of these approaches on a range of machine vision problems: adaptive background subtraction, estimating structure from motion and articulated body tracking.

CMU RI Defense: Modeling and Control Techniques for a Class of Mobile Robot Error Recovery Problems

Ravi Balasubramanian, Robotics Institute, Carnegie Mellon University
24 Jul 2006

Abstract: A robot's locomotion mode fails when its environmental contacts fail, a situation called a locomotion error. For example, a legged robot cannot move when its leg becomes trapped in a crevice, and a wheeled robot is handicapped when its wheels skid. How can a robot recover when its standard locomotion mode fails? One way is to utilize any remaining freedoms to move the robot to a situation where the robot's standard locomotion mode is again feasible. However, such unconventional motion is difficult, since the relationship between the robot's controls and its motion in a locomotion error is unclear. The uncertainty and the perceived "element of luck" in locomotion error recovery appears as a lack of structure, inducing operators to sometimes use random maneuvers which can worsen the predicament. This thesis proposes finding recovery strategies by exploiting the structure inherent to the robot's constrained mobility and environmental interaction in the locomotion error. A robot equipped with multiple locomotion modes, even some inefficient modes, can choose between them depending on the circumstances, ultimately contributing to robust mobility.

While robotic locomotion fails in many ways depending on the robot's design and the environmental interaction, this thesis finds novel recovery modes involving a combination of direct actuation and dynamically coupled actuation for two specific locomotion errors: first, a high-centered legged robot, where the robot's body is stuck on a rock and the robot's legs dangle in air; and second, a car trapped in a slippery pit. In the high-centered robot problem, we present a novel locomotion mode called ``legless locomotion'', that allows the robot to locomote simply by rocking its body back and forth using leg swing without feedback about the robot's body motions. We use experiments and computer simulation to identify legless locomotion's key elements and use simple models to derive an approximate control technique. In the stuck-car problem, we use computer simulation to find a control strategy involving wheel torques and an active-suspension that allows the car to roll out of the pit, while minimizing the work done and the perturbations to the car body and satisfying the contact constraints. Finally, we present a classification structure for locomotion errors based on environmental influence.

Further Details: A copy of the thesis oral document can be found at http://www.cs.cmu.edu/~bravi/thesis.pdf.

CMU RI proposal: Management of Distributed Schedules in Uncertain Environments using Simple Temporal Network Representations

Anthony Gallagher, Robotics Institute, Carnegie Mellon University

28 Jul 2006

Abstract: Cooperative agents working as a team to achieve a common goal need to coordinate their action so that a cohesive solution is obtained that achieves the team's objective. The practical constraints of many applications often times prevent any agent in the team from obtaining a global view of the team's problem, necessitating the development of distributed planning and scheduling techniques. While many researchers have focused on the distributed planning problem, multi-agent scheduling has received less attention. This research focuses on extending Simple Temporal Network (STN) based schedulers to operate in multi-agent domains with uncertain environments. We propose to develop pro-active techniques to increase the robustness of the agents's schedules to uncertain events, investigate reactive techniques to manage inconsistent information, and exploit the constraint conflict information provided by the STN to design a conflict-directed coordination approach.

Further Details: A copy of the thesis proposal document can be found at http://poma.ius.cs.cmu.edu/anthonyg/proposal.pdf.

Friday, July 21, 2006

IEEE News: NEW TAG USED TO MAP WHALE BEHAVIOR

In the latest issue of "IEEE Computer Graphics and Applications," researchers from the University of New Hampshire (USA) and the Stellwagen Bank National Marine Sanctuary discuss their findings on a new geospatial recording tag that tracks and visualizes Humpback whale behavior in deep areas of the ocean. The researchers hope the findings will minimize whale deaths and accidents with sailing and fishing equipment. A tag merges and compresses data in a flash drive before releasing itself from the animal and floating to the surface. This data is fed into a program, called GeoZui4D, which creates 3D images of the whale's swimming patterns. This pattern data is then fed into a program called TrackPlot, which creates temporal ribbons that show the whales' behavior, much like a map. Of the findings, the tag confirmed that whales do roll sideways during much of their time underwater (previously thought to only occur during feeding), and that they use different flukes strokes when descending and ascending. Read more, including image captures from the program (PDF): the link

Monday, July 17, 2006

NSC news: 日本科學技術會議目標:兒童守護機器人 20年後是否已實現?

作者:沈淑菁 現職:留日學人
文章來源:取材自2006/03/23朝日新聞
發佈時間:95.07.17

日本綜合科學技術會議(議長小泉純一郎首相)於3月22日發表於第3期科學技術基本計劃實施期間(2006年度〜2010年度)將重點實施的62項「戰略重點科學技術」。除了明確提出今後5〜20年欲達成之研究成果外,同時也首次公開各政府單位的完整配套政策方針。

該62項重點課題都將安全對策列為考量。該基本計劃欲達成的目標包含了活用情報通信技術與探測技術去探測化學物質及生化藥劑的反恐怖對策,以及開發守護街頭兒童的機器人等等。

See the full article.

Friday, July 14, 2006

IEEE news: PROTOTYPE DEVICE BLOCKS DIGITAL STILL AND VIDEO PHOTOGRAPHY

A device that can block digital camera function in a given area by looking for the reflectivity and shape of the cameras' charge couple devices (CCDs) has been developed by researchers at the Georgia Institute of Technology, who say the prototype uses off-the-shelf equipment -- camera-mounted sensors, lighting equipment, a projector and a computer -- to find and neutralize digital cameras. Researchers say the technology has commercial promise in protecting limited areas against clandestine photography and stopping video copying in larger areas such as theaters, which costs the film industry billions of dollars a year in lost revenue. The prototype uses visible light and two cameras to find CCDs, according to researchers, but a future systems might use invisible infrared lasers and photo-detecting transistors to scan for contraband cameras. The system would flash a thin beam of visible white light directly at the contraband camera's CCD, rendering recorded video unusable, researchers say. Read more: http://www.sciencedaily.com/releases/2006/06/060626092638.htm

IEEE news: SCIENTISTS DEVELOP COMPUTER THAT CAN READ PEOPLE'S MINDS

A computer that can identify and react to a human user's moods has been created by British and American scientists at the University of Cambridge in England and the Massachusetts Institute of Technology in the United States. The technology, which continues to be developed, allows a computer to be programmed to recognize facial expressions on actor's faces through a video camera. The researchers hope to use the technology to allow Web sites to create advertising that appeals to people’s moods. By linking a webcam with specific software, a person's image could be processed, allowing for the computer to analyze the accurate emotional state, and then send this data to a Web site. Other possible applications include dashboard computers and online teaching. Read more: http://abc.net.au/science/news/stories/s1671791.htm

CMU visit

The vehicle for 2007 Urban Challenge

The Red Team robots

John, Bob and Dave at CMU

Tuesday, July 11, 2006

News: Home as the Hub of Health Care

Many in government and society are banking on driving down the cost of health care with portable medical monitoring devices for the home and electronic medical records for the hospital and doctor's office. Much headway has been made on the hardware and software, but many policy issues remain. How to keep medical information from unauthorized eyes, who will pay for converting paper records to electronic files, and standards for the devices and the record-keeping systems themselves are just a few of the issues. Read on at http://bmsmail3.ieee.org:80/u/3349/41409275

News: 3-D Scanner Wins Designer the 2006 IEEE Presidents' Scholarship

A three-dimensional laser scanner built by high school sophomore Brandon Lee Reavis has won the budding engineer a US $10,000 scholarship from the IEEE Foundation. IEEE President-Elect Leah Jamieson presented Reavis with the 2006 IEEE Presidents' Scholarship for his project, "3-D Silhouette Laser Scanning: A Digital Reconstruction of Real-World Objects Into Point Clouds." Read more at http://bmsmail3.ieee.org:80/u/3354/41409275

Wednesday, July 05, 2006

My talk tomorrow (2006/7/6 Thu 10:30-) -- TOPIC CHANGE

Sorry about this.
Instead of vSLAM, I will talk about hardware stuff, such as:
- how to use 8051
- how to let 8051 communicate with PC
- how to use electronics parts, such as sensors
- how to solder (銲接)
- how to use the oscilloscope (示波器)
- etc..