Biotech Wizards Engineer Electronic Skin

Biotech wizards have engineered electronic skin that can sense touch, in a major step towards next-generation robotics and prosthetic limbs.

The lab-tested material responds to almost the same pressures as human skin and with the same speed, they reported in the British journal Nature Materials.

Important hurdles remain but the exploit is an advance towards replacing today’s clumsy robots and artificial arms with smarter, touch-sensitive upgrades, they believe.

The “e-skin” made by Javey’s team comprises a matrix of nanowires made of germanium and silicon rolled onto a sticky polyimide film.

The team then laid nano-scale transistors on top, followed by a flexible, pressure-sensitive rubber. The prototype, measuring 49 square centimetres (7.6 square inches), can detect pressure ranging from 0 to 15 kilopascals, comparable to the force used for such daily activities as typing on a keyboard or holding an object.

A different approach was taken by a team led by Zhenan Bao, a Chinese-born associate professor at Stanford University in California who has gained a reputation as one of the top women chemists in the United States.

Their approach was to use a rubber film that changes thickness due to pressure, and employs capacitors, integrated into the material, to measure the difference. It cannot be stretched, though.

The achievements are “important milestones” in artificial intelligence, commented John Boland, a nanoscientist at Trinity College Dublin, Ireland, who hailed in particular the use of low-cost processing components.

In the search to substitute the human senses with electronics, good substitutes now exist for sight and sound, but lag for smell and taste.

Touch, though, is widely acknowledged to be the biggest obstacle.

Even routine daily actions, such as brushing one’s teeth, turning the pages of a newspaper or dressing a small child would easily defeat today’s robots.

Bao added important caveats about the challenges ahead.

One is about improving the new sensors. They respond to constant pressure, whereas in human skin more complex sensations are possible.

This is because the pressure-sensing cells in the skin can send different frequencies of signal — for instance, when we feel something painful or sharp, the frequency increases, alerting us to the threat.

In addition, Bao warned, “connecting the artificial skin with the human nerve system will be a very challenging task”.

“Ultimately, in the very distant future, we would like to make a skin which performs really like human skin and to be able to connect it to nerve cells on the arm and thus restore sensation.

“Initially, the prototype that we envision would be more like a handheld device, or maybe a device that connects to other parts of the body that have skin sensation.

“The device would generate a pulse that would stimulate other parts of the skin, giving the kind of signal ‘my (artificial) hand is touching something’, for instance.”

In the future, artificial skin could be studded with sensors that respond to chemicals, biological agents, temperature, humidity, radioactivity or pollutants.

berkeley.edu



Applied Time and Motion

From the control system perspective, I find it interesting that we continue to model most industrial applications of motors with trapezoidal “time displacement” curves and PID (proprotional integral and derivative) tuning algortihms.  It seems that we should have better definitions for things after all the time and effort that goes into it.

It is important to see that the graphical representation of motion in “time displacement” plot is also a very literal representation of the mechanical aspect of motion.  The area under the curve is the mechanical work done by the system.

The acceleration leg of the trapezoid is the energy needed to bring the load from rest to a constant speed.  The acceleration profile is expressed simply as a scalar changing velocity that is increasing in value until the desired speed is reached.  The first derivative of velocity is acceleration, or the rate of increase of velocity over the unit change in time.

What might be of interest here is that the acceleration changes from a positive value to zero when the desired speed is reached.  Then acceleration is zero, because the system command will be for a constant speed with no acceleration.  And in PID type controls, the transition from positive acceleration to a zero acceleration invariably causes velocity overshoot.

Since everything can be mapped with respect to time, isn’t it more straightforward to consider the inflection points along the way and change the control methodology according to what is going on in the real world?

So coming up to the transition from accelerating to not accelerating, we know that the first derivative value is decreasing to zero.  Maybe our control strategy should be to switch from the velocity loop control and switch to the torque loop so we can decrease the current going to the motor.  This will soften the transition point and minimize overshoot without the use of PID control.

Torque control during fixed position control also has some intuitive benefits.  Any disturbance in position is countered by an opposing torque, or current control through the servoamplifier, to restore position.  This behavior can be embedded in the amplifier and does not require intervention from the controller, minimizing control system loading.

PID control, while enormously successful over the years, is a sort of averaging solution. We will apply gain values across a wide range of motion conditions in hope that they will work satisfactorily for all states over the time of the move .   But if we consider other possibilities, other strategies for control become possible.

This has been the goal of “adaptive gain” solutions which exist today and have evolved over the years as control technology has become cheaper and more powerful and the industry has acknowledged the weaknesses of PID control. By paying attention to the “inflection points” along the trajectory fo the motion, different control strategies are made possible that are simple, reliable and in some cases, more robust than what is possible with conventional control.

Robot Snake That Can Climb Trees & Spy

From the Biorobotics Lab at Carnegie Mellon University, a snake robot (Snakebot) demonstrates how it can climb a tree and look around.

Please keep in mind that this robot climbed a specific tree with a specific trunk width about 1 meter off of the ground. The researchers working to design, build and program these robots still have much work to do to get these bots to climb taller trees of various sizes and to navigate over branches and wires.  This may be especially useful in military tactics and surveillance

www.cmu.edu

Superior Feedback Performance in Telerobotics

WITTENSTEIN has perfected its control loading products to provide realistic force feedback for the telerobotics market. Utilizing compact design and unique electronic linking, sidestick systems from WITTENSTEIN offer revolutionary reliability and realism for operators.

WITTENSTEIN Aerospace & Simulation has been the control loading leader in the flight simulation market for more than a decade. The Company has taken its expertise and applied it to telerobotics, where a user controls an axis or entire vehicle remotely. WITTENSTEIN’s products provide the user with feedback of the remote axis through electrical linking and force control technology.

The main features of the sidestick systems for telerobotics are superior efficiency, compact design, and electric linking with force feedback. These result in smooth operator feel, no need for additional mechanical linkages or hydraulics, and a standard off-the-shelf system solution that utilizes standard wall-outlet power. The robust nature of the WITTENSTEIN systems allow for up to 10 axes per control module.

Sample areas of application for this technology include remote product testing for reasons due to environmental or equipment restrictions.

www.wittenstein-us.com

Robots and the Future – Part 2

Robotics researchers have been pushing the envelope for the last 30 years since the inception of “artificial intelligence”.  The basics of artificial intelligence programming is the modeling of human expertise and mimicking human behavior in a variety of circumstances.

One aspect of artificial intelligence gave rise to expert systems.  Complex systems like diesel locomotives are very difficult to repair because of the large number of parts operating together.  Human experience accumulated after years of working with diesel locomotives needed to be captured in order to prevent each generation from having to apprentice workers over long periods of time in order to learn how to troubleshoot these systems. So programmers in the early days of AI were employed to learn and program the diagnostic procedures developed by skilled workmen over many years.

These programs were very successful.  But in no way do they replace human intelligence and insight.  This is simply an example of subtlety in programming a specific area of human experience.  Speech recognition continues to be a challenge after decades of effort, limited to transcription applications and simple material handling instructions.

Another area that came up was large scale logistical mapping, another Expert System.  What is the most economical way to use airplanes to transport people around the US?  When you think of a large air carrier and the number of airplanes, flights, destinations and how they might be mapped together to get the best use out of the airplanes, it is a problem that is too large and complex for a single human to work with.  Enter the expert system programmer.

But in none of these cases can a computer program exceed the boundaries of it’s programming.  Can the autonomous Jeep get from it’s starting point to it’s destination?  Yes.  With many man-years of programming and a vast array of computing power, proper deployment of sensors and actuators, and a lot of stored energy.

Can the autonomous Jeep perform any other task?  No.  Regardless of the sophistication, the machine cannot exceed the boundaries of it’s programming.

Can we teach machines to learn?  So far, only in the most crude and rudimentary way.  But the course of the learning is again bounded by the programming.

And again, I will defer discussion of true intelligence or consciousness.

But what robotics can do to expand it’s usefulness is to mimic simple human tasking where it is cost effective and where the robot can “outproduce” or exceed the precision of a human.  Robotic welding, for example, has reached the point where a basic robot welding cell is less than $50,000.  So the cost of entry, the learning curve and complexity of implementing a welding robot cell in a small production facility is very reasonable.

Will robots be used in “human service” applications?  Sure.  ”Robot, vacuum my living room”  No sweat.  We can already do that with a Roomba only it doesn’t have voice recognition yet.  We have robots that can mow the grass in the front yard and avoid shrubs and trees.  Very cool.

Will we have robot servants like C3PO in Star Wars?  Hopefully more intelligent, C3PO was kind of dumb.  Simple tasks like serving a drink at a bar? Yes, that’s been done too, although it doesn’t have philosophical conversations with customers.

Will robots be able to provide basic care in hospitals and for the elderly?  Anything is possible. It will come down to how far we can push the envelope of programming, safety and return on cost.  Certainly we get robots to get a cold beer from the fridge.  But if the fridge is empty can it run out to the store and get us a six pack?

Not anytime soon.

Robotic Machining Cuts Part Lead-Time From Months To Days

Subtractive processes, often referred to as CNC machining, have not stood still in the rapid prototyping arena. Faster tool path generation is just one of the newer developments enabling machining to play a strong role in the rapid prototyping and direct digital manufacturing arena. Now, robotic machining has the potential to significantly affect the rapid casting arena, especially in the area of large castings. Tooling costs as well as lead times increase dramatically as parts get larger. The equipment needed to deal with the size and weight of extremely large parts becomes more rare and thus, more expensive. The larger the equipment used for these large parts, the slower it will operate due to its heavy physical characteristics. The most significant advantage that robotic machining seems to have is the fact that the robot moves independently of the work piece giving it the ability to feed as quickly on a large part as it does on a smaller, lighter part.

The US Department of Defense (DoD) has been seeking a way to reduce the cost of producing cast spare parts. The Advanced Technology Institute (ATI) currently leads several national collaborations that are developing advanced robotics capabilities and implementing both new and existing robotics technologies in response to the DoD’s need.

One collaboration is with the American Metalcasting Consortium (AMC). The ATI-managed AMC partner companies, like Clinkenbeard, are using robotics technologies to support legacy weapon systems; which could help meet the Defense Logistic Agency’s goal of dramatically shorter lead times for the production of legacy weapon systems parts. The patented Clinkenbeard® Toolingless Process proved that it could reduce lead times for military cast spare parts from six to twelve months to six to twelve days.

The results, according to ATI, also demonstrated that the Toolingless Process can reduce capital investment by as much as 35%, reduce individual parts cost by up to 20%, and improve cycle time by 25%.

Lead times often exceed a year because technical data may require reworking, including the development of a solid model of the part. But, even when a solid model is generated first, the Clinkenbeard process can supply a cast part in less than a month. The secret is computer-generated molds with no tooling.

The Toolingless Process consists of machining sand cores and molds, and is accurate. According to the company, this process can reduce the lead-time to obtain development castings by up to 90%. With this process, you can:

• eliminate the need for prototype tooling, depending on project requirements.
• make and test multiple design iterations during product development, from the simple to complex parts.
• reduce the cost of production tooling for one-of and small quantities.
• obtain accurate, prototype parts while large quantity tooling is made.
• eliminate tooling inventory.
• match exact production core materials and chemical levels so that prototype castings emulate production.
• incorporate engineering changes into high-volume production sand cores.

Clinkenbeard developed the sand machining process using CNC machining centers. By using robots with sand machining, company technicians can use the process on much larger molds and cores. Robotic technology will reduce the cost dramatically compared to the same expenditure for CNC machining centers.

Clinkenbeard
www.clinkenbeard.com

American Metal Consortium

http://amc.aticorp.org/

Defense Logistic Agency
www.dla.mil

Advanced Technology Institute (ATI)
www.aticorp.org

Iranian Robot Walks, Stands On One Leg

Researchers at Tehran University, in Iran, unveiled last month an adult-sized humanoid robot called Surena 2.

The initial press reports in Iran’s official news media didn’t include many details, saying only it could “walk like a human being but at a slower pace” and perform some other tasks, and there were questions about the robot’s real capabilities.

IEEE Spectrum obtained more information about Surena, as well as images and videos showing that the robot can indeed walk — and even stand on one leg.

Aghil Yousefi-Koma, a professor of engineering at the University of Tehran who lead the Surena project, tells me that the goal is to explore “both theoretical and experimental aspects of bipedal locomotion.”

The humanoid relies on gyroscopes and accelerometers to remain in balance and move its legs, still very slowly, but Yousefi-Koma says his team is developing a “feedback control system that provides dynamic balance, yielding a much more human-like motion.”

Surena 2, which weighs in at 45 kilograms and is 1.45 meter high, has a total of 22 degrees of freedom: each leg has 6 DOF, each arm 4 DOF, and the head 2 DOF. An operator uses a remote control to make the robot walk and move its arms and head. The robot can also bow. Watch:

Surena doesn’t have the agile arms of Hubo, the powerful legs of Petman, or the charisma of Asimo — but hey, this is only the robot’s second-generation, built by a team of 20 engineers and students in less than two years. A first version of the robot, much simpler, with only 8 DOF, was demonstrated in late 2008.

Yousefi-Koma, who is director of both the Center for Advanced Vehicles (CAV) and the Advanced Dynamic and Control Systems Laboratory (ADCSL) at the University of Tehran, says another goal of the project is to “to demonstrate to students and to the public the excitement of a career in engineering.”

Next the researchers plan to develop speech and vision capabilities and improve the robot’s mobility and dexterity. They also plan to give Surena “a higher level of machine intelligence,” he says, “suitable for various industrial, medical, and household applications.”

The robot was unveiled by Iranian President Mahmoud Ahmadinejad on July 3rd in Tehran as part of the country’s celebration of “Industry and Mine Day.” The robot is a joint project between the Center for Advanced Vehicles and the R&D Society of Iranian Industries and Mines.

Robots and the Future

In the field of Robotics, where is the line between between remote control, software control and autonomous control?  (No, I’m not going after the consciousness thing, it’s way too complicated)

Part of the problem may have to do with our use of the word “intelligence”.  We talk about the increasing “intelligence” of processors and particularly about the cost of “intelligent” control dropping to the point where it is suddenly economical to put a microcontroller together with a motor in order to achieve new levels of performance in either energy management or some other critical parameter.  Which opens new performance capability in robot design.

Increasingly, industrial robotics involve the use of vision systems to acquire information about the location and orientation of parts so that the robot system can interface smoothly to the “real world”.  If any of you have been to an industrial trade show and witnessed the Delta Robots making cookies, it is a very impressive sight to behold.  Incredible throughput and accuracy.  And that’s what it’s all about in industry. Higher productivity, improved product quality.

But where is the line between remote control and automatic control?  A remote manipulator for working in the nuclear industry, which was the big application that drove early robots, is a remote servo loop operating a series of servo motors and controls and powering mechanical systems, in order to do work that is dangerous to humans from a safe distance.  The DaVinci medical robot is a phenomenally improved version of the same thing.  A remote controlled robot, guided by direct haptic inputs from a surgeon, and with very sophistical tactile feedbacks, whose end effectors operate a variety of surgical instruments and actually increase the precision and speed with which doctors may perform certain procedures.

Is this a robot? Sure!

When we watch welding and painting robots making cars, we are watching decades of technology development in action.  There has been significant effort to improve the actuator hardware, and probably many man-years of software development to improve our description of the task and its safety and performance constraints in order to create not only reliable, but increasingly efficient machines to do the tasks that humans cannot compete with for productivity.  These are very sophisticated automatic applications, but certainly not autonomous.  The boundaries of the application and the programming for it are very finite.  Again, its about repetition, speed and accuracy.

And, yes, we call these robots, too.

But increasingly, there is discussion about the next frontier of robotics.  Where are the next big apps coming from?  Most of the big robotic companies in Japan and Europe are talking about personal service robots.  You can let your imagination run wild here.  Anything is possible. Certainly the service robot for NASA is interesting because it, again, follows the concept of doing tasks where it is difficult for humans to operate.

Is a Jeep that can be programmed to find a path and drive from one place to another autonomously a robot?  Yes, but we may be pushing the boundaries here just a bit.  These applications fall into the realm of Artificial Intelligence.  The programming and software languages for which were just being described for the first time about 30 years ago.  And at this point we are forced into the debate about what is intelligence.  In addition, are these systems which are capable of “learning” and what is learning exactly?  And more importantly, as all good science fiction movie watchers will ask, can a machine exceed it’s programming?  (See?  I didn’t even start on consciousness yet)

These are all serious considerations for the Future of Robotics which I will pick up further next week.

Giant Robotic Arm Simulates Driving a Ferrari

The hot-pink industrial arm whips you around while you sit in the driver’s seat


This image shows the robotic arm Ferrari simulator without a steering wheel attached. The simulator includes a force-feedback steering wheel and pedals.

Paolo Robuffo Giordano and colleagues at the Max Planck Institute for Biological Cybernetics in Tübingen, Germany, must really enjoy their jobs. Their CyberMotion Simulator is intended to realistically replicate the experience of driving a Ferrari without actually having to buy one.

Players sit in a cabin on a robot arm about 7 feet off the ground and drive a Ferrari F2007 car around a projected track. The robot arm, a type usually found in amusement parks, whips the driver around to simulate the Ferrari’s motion, according to IEEE Spectrum. You can hear the robot whine as the driver tries to turn at high speed.

The researchers wanted to use a robotic arm as a motion simulator with the goal of understanding how humans experience the sensation of motion. They figured an F1 racing game would be a good way to do it, IEEE Spectrum reports.They presented a paper on their design at the IEEE International Conference on Robotics and Automation this spring.

ieee.org

Pendulum-Tail Robot Climbs Vertical Walls In Seconds

Wielding two claws, a motor and a tail that swings like a grandfather clock’s pendulum, a small robot named ROCR (“rocker”) scrambles up a carpeted, 8-foot wall in just over 15 seconds – the first such robot designed to climb efficiently and move like human rock climbers or apes swinging through trees.

“While this robot eventually can be used for inspection, maintenance and surveillance, probably the greatest short-term potential is as a teaching tool or as a really cool toy,” says robot developer William Provancher, an assistant professor of mechanical engineering at the University of Utah.

His study on development of the ROCR Oscillating Climbing Robot is set for online publication this month by Transactions on Mechatronics, a journal of the Institute of Electrical and Electronics Engineers and American Society of Mechanical Engineers.

Provancher and his colleagues wrote that most climbing robots “are intended for maintenance or inspection in environments such as the exteriors of buildings, bridges or dams, storage tanks, nuclear facilities or reconnaissance within buildings.”

But until now, most climbing robots were designed not with efficiency in mind, only with a more basic goal: not falling off the wall they are climbing.

“While prior climbing robots have focused on issues such as speed, adhering to the wall, and deciding how and where to move, ROCR is the first to focus on climbing efficiently,” Provancher says.

One previous climbing robot has ascended about four times faster than ROCR, which can climb at 6.2 inches per second, but ROCR achieved 20 percent efficiency in climbing tests, “which is relatively impressive given that a car’s engine is approximately 25 percent efficient,” Provancher says.

The robot’s efficiency is defined as the ratio of work performed in the act of climbing to the electrical energy consumed by the robot, he says.

Provancher’s development, testing and study of the self-contained robot was co-authored by Mark Fehlberg, a University of Utah doctoral student in mechanical engineering, and Samuel Jensen-Segal, a former Utah master’s degree student now working as an engineer for a New Hampshire company.

The National Science Foundation and University of Utah funded the research.

ROCR is a Swinger that Claws Its Way to the Top

Other researchers have studied a variety of ways for climbing robots to stick to walls, including dry adhesives, microspines, so-called “dactyl” spines or large claws like ROCR’s, suction cups, magnets, and even a mix of dry adhesive and claws to mimic wall-climbing geckos.

Now that various methods have been tried and proven for robots to climb a variety of wall surfaces, “if you are going to have a robot with versatility and mission-life, efficiency rises to the top of the list of things to focus on,” Provancher says.

Nevertheless, “there’s a lot more work to be done” before climbing robots are in common use, he adds.

Some previous climbing robots have been large, with two to eight legs. ROCR, in contrast, is small and lightweight: only 12.2 inches wide, 18 inches long from top to bottom and weighing only 1.2 pounds.

The motor that drives the robot’s tail and a curved, girder-like stabilizer bar attach to the robot’s upper body. The upper body also has two small, steel, hook-like claws to sink into a carpeted wall as the robot climbs. Without the stabilizer, ROCR’s claws tended to move away from the wall as it climbed and it fell.

The motor drives a gear at the top of the tail, causing the tail to swing back and forth, which propels the robot upward. A battery is at the end of the tail and provides the mass that is necessary to swing the robot upward.

“ROCR alternatively grips the wall with one hand at a time and swings its tail, causing a center of gravity shift that raises its free hand, which then grips the climbing surface,” the study says. “The hands swap gripping duties and ROCR swings its tail in the opposite direction.”

ROCR is self-contained and autonomous, with a microcomputer, sensors and power electronics to execute desired tail motions to make it climb.

Provancher says that to achieve efficiency, ROCR mimics animals and machines.

“It pursues this goal of efficiency with a design that mimics efficient systems both in nature and manmade,” he says. “It mimics a gibbon swinging through the trees and a grandfather clock’s pendulum, both of which are extremely efficient.”

The study says: “The core innovations of ROCR – its energy-efficient climbing strategy and simple mechanical design – arise from observing mass shifting in human climbers and brachiative [swinging] motion in animals.”

Simulating and Testing a Climbing Robot

Before testing the robot itself, Provancher and colleagues used computer software to simulate ROCR’s climbing, using such simulation to evaluate the most efficient climbing strategies and fine-tune the robot’s physical features.

Then they conducted experiments, varying how fast and how far the robot’s tail swung, to determine how to get the robot to climb most efficiently up an 8-foot-tall piece of plywood covered with a short-nap carpet.

The robot operated fastest and most efficiently when it ran near resonance – near the robot’s natural frequency – similar to the way a grandfather clock’s pendulum swings at its natural frequency. With its tail swinging more slowly, it climbed but not as quickly or efficiently.

The researchers found it achieve the greatest efficiency – 20 percent – when the tail swung back and forth 120 degrees (or 60 degrees to each side of straight down), when the tail swung back and forth 1.125 times per seconds and when the claws were spaced 4.9 inches apart.

When the tail swung at two times per second, it was too fast and ROCR jumped off the wall, and was caught by a safety cord so it wasn’t damaged.

Provancher says the study is the first to set a benchmark for the efficiency of climbing robots against which future models may be compared. He says future work will include improving the robot’s design, integrating more complex mechanisms for gripping to walls of various sorts, such as brick and sandstone, and investigating more complex ways of controlling the robot – all aimed at improving efficiency.

“Higher climbing efficiencies will extend the battery life of a self-contained, autonomous robot and expand the variety of tasks the robot can perform,” he says.

« Previous Page — Next Page »