TiLR: A telepresence robot in action

RoboDynamics is hoping that there is a large market for telepresence robots. The idea of creating a remote robot body for people who need to attend meetings but cannot be at location in person is not new. However, I don't know of anyone who has successfully marketed a telepresence robot. RoboDynamics hopes to be the first with their TiLR robot which is currently undergoing trials.

You can see the TiLR robot in the photo at the top left (click the image for a larger view.) Clearly, the robot has a small LCD screen which displays an image of the user and a small camera with pan-tilt capabilities. The camera can be controlled remotely, of course. Needless to say, the robot has a bidirectional audio and video link so that its user can interact with the people at the remote location.

The following promotional video explains using a simple cartoon how the company believes their telepresence robot can be used to add value to a company.

Finally, the following Mahalo Daily video showcases the actual TiLR robot and its control software.

IntellaSys SEAforth 40C18: A low-power 40-core scalable embedded array processor

It is not likely that we will have advanced robots in our homes unless a major breakthrough in low-power, high-performance embedded processors occurs. IntellaSys is close to this breakthrough with the release of the SEAforth 40C18 embedded processor which boasts 40 cores and operates at low power. In fact, the company claims that SEAforth uses 28 times less power and runs 240 times faster than other competing systems.

Each of the cores is a standalone microprocessor with its own RAM and ROM. All cores together can work asynchronously to process data in parallel. If necessary, the cores can communicate via dedicated ports. Best of all, the software developers need not worry about the synchronization because IntellaSys has a hardware solution that takes care of it.

With 40 nodes to work with, designers can dedicate groups of them to specific tasks such as FFT and DFT algorithms. The result is a tightly coupled, extremely versatile user-defined group of dedicated processors assigned to specific tasks. Some can be doing highly compute-intensive audio processing, while others handle wireless interfaces, external memory, and user interface functions. And since each core has its own ROM and RAM, there is less need to go to external memory

The new computer is easily programmable using the VentureForth IDE which includes debugging tools and a simulator. VentureForth is available for both Microsoft Windows and Linux.

Opportunity's new 2-year mission

NASA's twin Mars rovers Spirit and Opportunity have done a fantastic job exploring the surface of Mars since 2004. Their initial mission was meant to last just 3 months but both rovers have continued to operated for 4 years on and there is no stopping them. In fact, Opportunity is about to go on its newest mission which is a 2-year long trek to a massive crater named Endeavour.

The BBC has a longer story on this new mission. Opportunity will have to navigate 11km of rough terrain to reach the crater; engineers suggest that the robot will traverse about 100 meters daily during the summer months when sunlight is plentiful for supplying power to recharge the rover's batteries. NASA engineers have uploaded new software on the robot that will allow it to navigate with more autonomy while the orbiting Mars Reconnaissance Orbiter will create maps of the planet's surface ahead of time to guide the rover along the way.

What a great achievement for NASA to build and deploy a multi-robot system in one of the most difficult environments known to man. These robots must complete their mission with as little guidance from human operators as possible with little hope for rescue if things go really bad. I will definitely be watching as Opportunity executes its new 2-year long trek towards Endeavour. When it completes its mission, it will be a new record for a team of robots working together to complete a complex mission.

OOPSMP: An Object-Oriented Programming System for Motion Planning

Motion planning is an essential component for any robot and there is a very large body of literature describing different methods some of which work better than others. Lydia Kavarki's research group at Rice University continues to update their state of the art Object-Oriented Programming System for Motion Planning which is free to download and use. This is how the group describes the motion planning system in one sentence:

OOPSMP is a package for motion planning that is easy to extend, robust, and efficient. It can be used for motion planning research or as a teaching tool.

The software is good enough for use in teaching but also research. It is object oriented and it can be easily extended. For those who are not interested in the low-level details, OOPSMP is easy to use via a high-level interface that allows users to specify the planning problem in XML, select the algorithm to use, and let the system do all the work.

Professor Kavraki is well known in the robotics and planning communities for her work on Probabilistic Roadmaps (PRMs). Kavarki introduced PRMs more than 10 years ago in 1996. Simply put, the method works as follows (source: Wikipedia)

The basic idea behind PRM is to take random samples from the configuration space of the robot, testing them for whether they are in the free space, and use a local planner to attempt to connect these configurations to other nearby configurations. The starting and goal configurations are added in, and a graph search algorithm is applied to the resulting graph to determine a path between the starting and goal configurations.

If you are building a robot and you have a motion planning problem then you should consider downloading a copy of OOPSMP. The software is free to use for non-commercial applications.

Free robotics courses from Stanford Engineering

Stanford University has started a new website with 10 free courses in Computer Science, Robotics, and Linear Systems and Optimization. The new Stanford Engineering Everywhere (SEE) initiative is a great opportunity for attending lectures by some of the top researchers and scientists all from the convenience of our own homes. Among the courses listed, there is a course in robotics and a course in machine learning.

The robotics course is taught by Oussama Khatib who is well known for his contributions in the field; he has published some 200 papers during 25 years of research and he is still going strong. The focus of Khatib's work these days is in human-centered robotics and haptic human-robot interaction.

The machine learning course is taught by Andrew Ng whose robots we have mentioned numerous times in this blog. If you don't remember, most recently we wrote about the latest iteration of Ng's autonomous RC helicopter that learns the autorotation maneuver using reinforcement learning. If you want to learn how they did it then you should follow his course on SEE. I have embedded below Ng's first lecture which gives a nice overview of machine learning (the first few minutes are boring but the lecture becomes very interesting after that.)

The US army wants more robot soldiers fast

RF Design magazine has published an interesting article about the US army's desire to replace 30% of its ground forces by autonomous systems in a little more than 10 years from now. The article's author interviews assistant professor Bill Smart and his student Doug Few who are two of the researchers working towards achieving the army's goal.

According to the article, Smart and Few are working on developing semi-autonomous robot soldiers that will let humans make the call before releasing death upon the enemy. For obvious reasons the army does not want to develop fully autonomous robot soldiers since in such a case there is nothing to prevent friendly fire. Moreover, Few wants to utilize a Wii remote in order to enable human soldiers to give robot soldiers commands using natural gestures.

This type of technology may be created at first for a not so friendly application but if successful it should find its way into consumer robots. At this stage, it is next to impossible to instruct a home robot (not that any such robots really exist today) to perform a task using natural hand gestures or even voice commands.

Serial versus parallel processing demonstrated

You may have already seen this video (I have received a link to it from 3 different friends on 3 different continents,) but if you haven't then you should. Adam Savage and Jamie Hyneman co-hosts of the very popular Mythbusters TV show put together a clever demonstration to show the vast speed-up that can be gained from parallel processing over serial processing. The demonstration took place at NVIDIA's NVISION 08 and is known under the title "CPU vs. GPU." Needless to say, the benefits of parallel processing can only be realized if one can devise appropriate algorithms that take full advantage of parallel processing hardware such as a GPU. Needless to say, if such an algorithm exists, then...well just watch the video below.

Keeping a ball balanced in real-time using a touch screen

The researchers at the Real-Time Systems Laboratory (RETIS lab) of the Scuola Superiore Sant'Anna in Pisa have created a nice system that manages to keep a steel ball balanced in the middle of a touch screen. The system consists of a Flex board with a Microchip dsPIC DSC and a couple of servo motors other than the touch screen, of course. I don't know much about electronics but I know that real-time control is hard. The video below shows the ball and plate system in action. Very cool!

ScanRobot: An innovative book scanning robot

There is no doubt that today there is a huge interest in digitizing large collections of books many of which were written well before digital computers were an important tool for authors. But digitizing a book is a difficult task. This is a job well suited for a robot. Treventus Mechatronics has a very nice solution to the book scanning problem with their ScanRobot offering. This innovative robot can scan any book at a rate of 2400 pages per hour. The ScanRobot uses patented technology to scan and turn pages easily and naturally without damaging the book. The video below shows the automatic book scanner ScanRobot in action.