Friday, August 31, 2007

Invention: UAV swarms

* 13:12 28 August 2007
* news service
* Justin Mullins

The skies above future battlefields are likely to be filled with unmanned aerial vehicles (UAVs) monitoring the action and homing in on enemy targets. But how do you control a sky full of UAVs, particularly when communications links with the ground are patchy?

Perhaps you don't have to. Yossi Ben-Asher and colleagues at the Technion-Israel Institute of Technology in Haifa, Israel, say UAVs can control themselves instead.

Their idea is for a swarm of UAVs to create its own communications network and swap the information needed to calculate flight paths and avoid collisions. This not only eliminates the need for constant communication with the ground, it can also be used to make the aircraft swarm together, like a flock of birds.

This swarming behaviour can then be used to mount more effective attacks, the team says. The UAVs could monitor a target and decide for themselves how best to attack it, based on the position of each aircraft and the weapons it is carrying.

The patent application stops short of explaining how the final decision to engage is taken. Presumably somebody on the ground has to give the go-ahead. But how long before technology like this makes the military think it can dispense with even this step?

Read the full swarming UAVs software patent application.

Thursday, August 30, 2007

News: Autonomous vehicle finds way without GPS

In the European Land Robot Trial, a vehicle of the university of the German armed forces (Universitaet der Bundeswehr) is the winner in the speed discipline. In comparison the Darpa Grand Challenge, the European contest demands more autonomy, participants say.

The vehicle used by the university's team was a Volkswagen Tuareg, the same type that already has won last year's Grand Challenge. Basically, the vehicle even was equipped very similarly, but in contrast to the Darpa Grand Challenge where the vehicles drive along a track that connects waypoints, defined by GPS navigation, the European Land Robot Trial does not allow the intensive use of satellite navigation.

EE Times Europe: Autonomous vehicle finds way without GPS - Link
European Land-Robot Trial (ELROB) Website - Link
Wikipedia: ELROB - Link

Wednesday, August 29, 2007

New devices promise touchy-feely computing

Full content

Is it possible to "feel" an object while being in another location? This is a question addressed by several technologies on show at the SIGGRAPH 2007 computer conference in San Diego, California, US, earlier this month.

Box of balls
Gravity Grabber, this prototype device consists of two tube-shaped attachments that fit over a person's thumb and forefinger. Each tube has a pair of motors that sit on top and connect to a belt that wraps around the tip of the finger. The motors can pull on the belt to produce the feeling of touching an object, and can produce the feeling of holding something heavy by pulling more tightly.

Gravity Grabber in action

Monday, August 27, 2007

ICAR2007 : Robot positioning using structured light patterns suitable for self calibration and 3D tracking

Title : Robot positioning using structured light patterns suitable for self calibration and 3D tracking

Author : Kasper Claes, Herman Bruyninckx

Abstract : We study the control of a robot arm with an eye-in-hand camera, using structured light with a projector in a fixed but unknown position. Based on the theory of perfect maps, we propose a deterministic method to generate patterns independent of the relative 6D orientation between the camera and the projector, with any alphabet size and minimal Hamming distance between the codes. This independence makes self calibration using a known robot motion possible, and removes the need for less robust calibration grids. Patterns that are large enough for controlling a robotic arm are generated with error correcting capabilities. The experiments show the controlled motion of an industrial robot to a chosen position near a non-planar object, using a projector-camera calibration. This provides a wider baseline and hence more robustness than structure from motion.

Lab Meeting 27 August (Chihao): Demonstration of Acoustic Localization in PAL2

I will show the acoustic localization system in PAL2.
This system could find the direction of sound audio source even if the source is moving.

Sunday, August 26, 2007

Lab Meeting 27 August (Atwood): Recognition by Local Feature

I will talk about two works on recognition by local features and my recent experiments

Jerry Jun Yokono, Tomaso Poggio. A Multiview Face Identification Model With No Geometric Constraints. In Proceedings of the 7th International Conference on Automatic Face and Gesture Recognition (FGR06)

Ariadna Quattoni, Michael Collins and Trevor Darrel. Conditional Random Fields for Object Recognition. In Advances in Neural Information Processing Systems 17 (NIPS 2004), 2005.

Saturday, August 25, 2007

Lab Meeting 27 August (fish60): Speeding up Moving-Target Search


In this paper, we study moving-target search, where an agent has to catch a moving target . The agent does not necessarily know the terrain initially but can observe it within a certain sensor range around itself.
It uses the strategy to always move on a shortest presumed unblocked path toward the target. We study how the agent can find such paths faster by exploiting the fact that it performs A* searches repeatedly.
To this end, we extend Adaptive A* to moving-target search and demonstrate experimentally that the resulting MT-Adaptive A* is faster than isolated A* searches.


News: LucidTouch - A See-Through Mobile Device

Softpedia - Windows Aero Is Old News - Link
Engadget - Microsoft R&D strikes again: the LucidTouch - Link

Wednesday, August 22, 2007

News: iRobot's new Roomba 560, 530: totally redesigned vacuumbots

iRobot hasn't released a new, proper Roomba servantbot since late 2004's Discovery (Scooba, of course, excepted). Today that changes, with the 560 and 530 we caught wind of back in May. There's a lot going on here, so we'll start at the top:
  • iRobot is touting these new Roombas as being "totally redesigned" from the ground up.
  • The new vacuum system is supposedly 100% more effective at picking up the grime in your floors. Really makes the old model suck. Thanks, we'll be here all week.
  • The 560 features 2.4GHz RF that activates its new Virtual Wall Lighthouse system (we'll get to that in a sec. The 530 doesn't have RF.
  • Scheduling is now built into the robot. Friggin FINALLY!
  • The new Roombas have IR sensors built into the bumper, and have a second, slower running speed. When Roomba is approaching an object, it slows down so as not to bump it so hard.
  • Lighthouses are Roomba's new RF-based progress tracking system. Place one in each doorway, and they switch on via RF when your Roomba does. When Roomba's done with a room, it won't re-enter until the next time it runs.
  • The battery is still removable, but you have to unscrew the bottom plate to get it out. Oh, and now you can add faceplates.
  • The system is far better sealed off, so most of the crap you're vacuuming up supposedly won't actually get inside the robot anymore. Extra bonus: that black arc between the bumper and buttons is a handle that lies flush.
  • The 560 goes for $350, the 530 goes for $300 -- snag 'em both starting now.
engadget: iRobot's new Roomba 560, 530 - Link
iRobot Corporation - Link

News: The DIY Segway

A tenacious combination of high school and MIT students have put together their very own DIY Segway, which would likely have Dean Kamen rolling in his grave... if he were actually dead. The team of ingenious do-it-yourself'ers put together a fairly workable Segway knock-off for less than a thousand dollars that can keep pace with the actual model (11 MPH versus the real-deal's 12.5). By using off-the-shelf supplies, like a Machine Science XBoard microcontroller, pieces of Lexan for the steering mechanism, and light, cheap aluminum for the frame, the gang was able to assemble a fairly stripped down version of the scooter. Apparently, the kids are still working out the kinks, but if this keeps up, Kamen and crew might have to start watching their backs. Check the video after the break to see the still-shaky personal transporter in action. - [via] Link

Friday, August 17, 2007

News: Nao bot chosen as AIBO RoboCup replacement

In a rather strange move, French start-up Aldebaran, and their Nao humanoid robot, has been chosen as the replacement competitor for the discontinued Sony AIBO. The two-legged Nao will replace the four-legged AIBO in the Standard Platform League (formerly the Four-Legged League) of RoboCup, the popular robotic soccer competition. It seems something of a curious choice, given the fact that Aldebaran hasn't even shipped a product yet. - [via] Link

Aldebaran Robotics - Link

Wednesday, August 15, 2007

Weight Loss Robot Will Shame You Into Eating Less

Hey, chubster. Do you have trouble keeping a healthy diet? Would having a friendly, diet-focused robot in your kitchen help? No? Well, don't tell the guys at the MIT Media Lab that, because that's exactly what they've been cooking up. It's an in-home robot that'll make eye contact with you while it tracks your eating habits and shuns you for constantly nibbling on Doritos rather than carrot sticks. It's like a nagging spouse, but it doesn't have the personal investment of having to see you naked. I'm not sure if that's a plus or a minus in its favor.


Tuesday, August 14, 2007

[VASC Seminar Series ]Video Summarization for PVRs

Speaker: Ajay Divakaran, MERL

Time: Wednesday August 15, 12-1pm


We present the world's first highlights-playback-capable Hard Disk Drive
(HDD)-enhanced DVD recorder. It automatically detects highlights in sports
video by detecting portions with a mixture of the commentator's excited
speech and cheering, using Gaussian Mixture Models (GMM's) trained using
the MDL criterion. Our computation is carried out directly on the MDCT
coefficients from the AC-3 coefficients thus giving us a tremendous speed
advantage. Our accuracy of detection of sports highlights is high across a
variety of sports. Our user-study shows that viewers like the new
functionality even if it makes mistakes. Finally, we propose
genre-independent temporal segmentation of non-sports content using
computationally inexpensive audio-visual features. Such segmentation
enables "smart skipping," from one semantic unit to another.

project link

[Book] Face Recognition

Title: Face Recognition

About the Book:

… this book will serve as a handbook for students, researchers and practitioners in the area of automatic (computer) face recognition and inspire some future research ideas by identifying potential research directions. The book consists of 28 chapters, each focusing on a certain aspect of the problem. Within every chapter the reader will be given an overview of background information on the subject at hand and in many cases a description of the authors' original proposed solution. The chapters in this book are sorted alphabetically, according to the first author's surname. They should give the reader a general idea where the current research efforts are heading, both within the face recognition area itself and in interdisciplinary approaches.

the full text of this book

Sunday, August 12, 2007

Lab Meeting 13 August (Yu-Chun): Emotional Architecture for the Humanoid Robot Head ROMAN

2007 IEEE International Conference on Robotics and Automation

Author: Jochen Hirth, Norbert Schmitz and Karsten Berns

Humanoid robots as assistance or educational robots is an important research topic in the field of robotics. Especially the communication of those robots with a human operator is a complex task since more than 60% of human communication is conducted non-verbally by using facial expressions and gestures. Although several humanoid robots have been designed it is unclear how a control architecture can be developed to realize a robot with the ability to interact with humans in a natural way. This paper therefore presents a behavior-based emotional control architecture for the humanoid robot head ROMAN. The architecture is based on 3 main parts: emotions, drives and actions which interact with each other to realize the human-like behavior of the robot. The communication with the environment is realized with the help of different sensors and actuators which will also be introduced in this paper.


Lab Meeting 6 August (Jim): Fast Replanning for Moving-Target Search

Talk by Sven Koenig

I will try to talk about this part: Fast Replanning for Moving-Target Search

MT-Adaptive A*, an incremental variant of A* that updates the heuristics between searches. MT-Adaptive A* is faster than isolated A* searches and, in many situations, also D* Lite, a state-of-the-art incremental variant of A*.

See here for more information
(Speeding up Moving-Target Search)

Thursday, August 09, 2007

Talk by Sven Koenig; Tuesday August 7

Computer Science Department, University of Southern CaliforniaPICTURE: See
DATE: August 7, TuesdayTIME: 10:30am-11:30am
LOCATION: Newell Simon 1305

An Overview of two of currentresearch projects on path planning for video games, namely, any-anglepath planning and fast replanning for moving-target search.

Any-Angle Path Planning:
Theta*, a variant of A* that propagates information along grid edges without constraining the paths to grid edges. Theta* is simple, fast and finds short and realistic looking paths.

Fast Replanning for Moving-Target Search:
MT-Adaptive A*, anincremental variant of A* that updates the heuristics between searches. MT-Adaptive A* is faster than isolated A* searches and, in many situations, also D* Lite, astate-of-the-art incremental variant of A*.

Additional information about Sven can be found on his web pages:

Wednesday, August 08, 2007

[Master Thesis] Incremental learning for Adaptive Visual Place Recognition in Dynamic Indoor Environments

Jie Luo,
Department of Computer and Systems Sciences,
Stockholm University / Royal Institute of Technology

Vision-based place recognition is a desirable feature for an autonomous mobile system. In order to work in realistic scenarios, a visual recognition algorithm should have two key properties: robustness and adaptability. This thesis focuses on the latter, and presents a discriminative incremental learning approach to place recognition. We propose a solution based on incremental extensions of support vector machine classifier. Since computational and memory efficiency are crucial for mobile robot platforms aim to continoulsy work in real-world settings, we put emphasis on these properties. We use a recently introduced memory-controlled incremental technqiue, which allows to control the memory requirements as the system updates its internal representation. At the same time, it preserves the recognition performance of the batch algorithm and runs online. In order to assess the method, we acquired a database capturing the intrinsic variability of places over time. Extensive experiments show the power and the potential of the approach.

Moreover, the feature space of SVM consists of both global feature - Composed Receptive Field Histograms, and local features - Harris-Laplace Detector and SIFT Descriptor.


Tuesday, August 07, 2007

CVPR 07 Awards

Longuet-Higgins Prize

* J. Shi and J. Malik, Normalized Cuts and Image Segmentation
* E. Osuna, R. Freund, and F. Girosi, Training Support Vector Machines: An Application to Face Detection

Best Paper Prize

* B. Leibe, N. Cornelis, K. Cornelis, and L. van Gool, Dynamic 3D Scene Analysis from a Moving Vehicle

Best Paper Prize Runner Ups

* A. Levin, A. Rav Acha, and D. Lischinski, Spectral Matting
* O. Tuzel, F. Porikli, and P. Meer, Human Detection via Classification on Riemannian Manifolds

Best Student Paper Prize

* Y. Li, H. Ai, T. Yamashita, S. Lao, and M. Kawade, Tracking in Low Frame Rate Video: A Cascade Particle Filter with Discriminative Observers of Different Lifespans

CNN news: Concept car gets testy with drunk drivers

A new concept car with breathalyser-like detection systems may provide even greater traction for Japanese efforts to keep impaired drivers off the road.

Nissan's alcohol-detection sensors check odor, sweat and driver awareness, issuing a voice alert from the navigation system and locking up the ignition if necessary.

Odor sensors on the driver and passenger seats read alcohol levels, while a detector in the gear-shift knob measures the perspiration of the driver's palm when starting the car.
But Nissan's car includes a mounted camera that monitors alertness by eye scan, ringing bells and issuing a voice message in Japanese or English if a driver should pull over and rest.

See the full article.

Sunday, August 05, 2007

Lab Meeting 6 August (Leo): Multiple-Model Estimation with Variable Structure

Multiple-Model Estimation with Variable Structure

author: Xiao-Rong Li,Yaakov Bar-Shalom

Existing multiple-model (MM) estimation algorithms have a fixed structure, i.e. they use a fixed set of models. An important fact that has been overlooked for a long time is how the performance of these algorithms depends on the set of models used. Limitations of the fixed structure algorithms are addressed first. In particular, it is shown theoretically that the use of too many models is performance-wise as bad as that of too few models, apart from the increase in computation. This paper then presents theoretical results pertaining to the two ways of overcoming these limitations: select/construct a better set of models and/or use a variable set of models. This is in contrast to the existing efforts of developing better implementable fixed structure estimators. Both the optimal MM estimator and practical suboptimal algorithms with variable structure are presented. A graph-theoretic formulation of multiple-model estimation is also given which leads to a systematic treatment of model-set adaptation and opens up new avenues for the study and design of the MM estimation algorithms. The new approach is illustrated in an example of a nonstationary noise identification problem.


Lab Meeting 6 August (Yu-Hsiang) : Unsupervised Activity Perception by Hierarchical Bayesian Models

Author: Xiaogang Wang, Xiaoxu Ma, Eric Grimson

Abstract :
We propose a novel unsupervised learning frameworkfor activity perception. To understand activities in complicatedscenes from visual data, we propose a hierarchicalBayesian model to connect three elements: low-level visualfeatures, simple “atomic” activities, and multi-agent interactions.Atomic activities are modeled as distributions overlow-level visual features, and interactions are modeled asdistributions over atomic activities. Our models improve existinglanguage models such as Latent Dirichlet Allocation(LDA) and Hierarchical Dirichlet Process (HDP) by modelinginteractions without supervision. Our data sets arechallenging video sequences from crowded traffic sceneswith many kinds of activities co-occurring. Our approachprovides a summary of typical atomic activities and interactionsin the scene. Unusual activities and interactions arefound, with natural probabilistic explanations. Our methodsupports flexible high-level queries on activities and interactionsusing atomic activities as components.


Lab Meeting 6 August (Vincent): 3D Active Appearance Modeling

In this talk, I will go through the details of the following components :

1. 2D+3D AAM (CMU's work)

2. My proposed approach.

Friday, August 03, 2007

NewScientist: Sharing a joke could help man and robot interact

* 01 August 2007
* news service
* Michael Reilly

A MAN walks into a bar: "Ouch!" You might not find it funny, but at least you got the joke. That's more than can be said for computers, which, despite radical advances in artificial intelligence, remain notably devoid of a funny bone.

Previously AI researchers have tended not to try mimicking humour, largely because the human sense of humour is so subjective and complex, making it difficult to program.

Now Julia Taylor and Lawrence Mazlack of the University of Cincinnati in Ohio have built a computer program or "bot" that is able to get a specific type of joke - one whose crux is a simple pun. They say this budding cyber wit could lend a sense of humour to physical robots acting as human companions or helpers, which will need to be able to spot jokes if they are to be accepted and not just annoy people. The bot is also teasing apart why some people laugh at a joke, such as the one above, when most just groan.

See the full article.

Thursday, August 02, 2007

Murata boy - bicycle riding robot

Height: Approx. 50cm
Weight: Approx. 5kg
Place of Birth: Kyoto, Japan
Zodiac sign: Libra
Hobby: Bicycling
Life goal: Riding around the world
Motto: When you fall of a bicycle, get right back on.

Official site: Link
CM: Link

Talk this Friday: Direct care workers in long-term care facilities in the United States

This talk could be related to our carebot and HRI projects. -Bob

講者:美國南卡大學教授 Rita Jing-Ann Chou, Ph.D.
間:2007年8月3日(五) 10:00-12:00

隨著老年人口的快速增加,迎合銀髮族需求的「長期照護」在美國成為一個最迫切需要的服務之一。這些長青老人,一方面他們因為較易受到傷害,不適合單獨居住;另方面,他們也並非脆弱到需要 24小時的醫護照料。養護中心(assisted living)的設計因為正符合上述需求,近年在美國蓬勃發展而且成為長期照護的主要型態之一。養護中心的照料結合了住宿與廣泛的個人健康醫療設施和協助服務。而這些服務大部分是由直接護理工作者( direct care worker)一肩挑起,因此,他們對於住在養護中心的人而言,扮演非常重要的角色。不幸的是,如同在其他型態的居住性長期治療處所,養護中心也因為人員高度異動,而困擾不已,有時一年的離職率竟可高達 200%。


3.研究結果對養護中心及對於社會工作及社會政策之意涵為何 ?


Wednesday, August 01, 2007

DIY: Make your own multitouch displays and software apps

Multitouch display technology has been gaining a lot of visibility recently, both with Microsoft's Surface interactive tabletop displays and the iPhone's slick gesture-based interface. Ignoring some of the particulars, the guts behind this technology is relatively simple, and you can make you own multitouch interactive display for little over the cost of a projector (the most expensive part of this setup).

An acrylic panel is edge lit with infrared leds. When your finger comes in contact with the acrylic, it scatters infrared light out the back where it is visible via infrared camera. As long as nothing is touching the acrylic, very little of the light escapes, instead just reflecting around inside. Image processing takes care of detecting tips of fingers and relaying their location to application software. Since the camera "reads" the whole display in parallel, it is easy to detect multiple fingertips at once, even those belonging to multiple users. All this sensing goes on in the infrared spectrum, leaving us free to utilize the visible spectrum to display interactive software.

As you might imagine, there's a lot of software that translates the infrared fingerpresses that the webcam sees into a usable operating system or application interface. You're not entirely on your own with this. There's an image processing library called Touchlib which will handle passing your C++ app screen touch events. The community of Touchlib developers seem to be pretty active and have put together a number of open source, sample applications which can help you get started.

Full Article - Link
DIY Multitouch Dsiplay @ Instructables - Link
Touchlib - Link
Natural User Interface Group (multitouch developer community) - Link