Humans are able to find objects in their surroundings and detect some of their properties simply by touching them. While this skill is particularly valuable for blind individuals, it can also help people with no visual impairments to complete simple tasks, such as locating and grabbing an object inside a bag or pocket. 

Researchers at Massachusetts Institute of Technology (MIT) have recently carried out a study aimed at replicating this human capability in robots, allowing them to understand where objects are located simply by touching them. Their paper, pre-published on arXiv, highlights the advantages of developing robots that can interact with their surrounding environment through touch rather than merely through vision and audio processing.

“The goal of our work was to demonstrate that with high-resolution tactile sensing it is possible to accurately localize known objects even from the first contact,” Maria Bauza, one of the researchers who carried out the study, told TechXplore. “Our approach makes an important leap compared to previous works on tactile localization, as we do not rely on any other external sensing modality (like vision) or previously collected tactile data related to the manipulated objects. Instead, our technique, which was trained directly in simulation, can localize known objects from the first touch which is paramount in real robotic applications where real data collection is expensive or simply unfeasible.”

As it is trained in simulations, the technique devised by Bauza and her colleagues does not require extensive data collection. The researchers initially developed a framework that simulates contacts between a given object and a tactile sensor, thus assuming that a robot will have access to data about the object it is interacting with (e.g., its 3-D shape, properties, etc.). These contacts are represented as depth images, which show the extent of an object’s penetration into the tactile sensor.

Subsequently, Bauza and her colleagues used state-of-the-art machine-learning techniques for computer vision and representation learning to match real tactile observations gathered by a robot with the set of contacts generated in simulation. Every contact in the simulation dataset is weighed depending on the likelihood that it matches the real or observed contact, which ultimately allows the framework to attain the probability distribution over possible object poses.

“Our method encodes contact, represented as depth images, into an embedded space, which greatly simplifies computational cost allowing real-time execution,” Bauza said. “As it can generate meaningful pose distributions, it can be easily combined with additional perception systems. In our work, we exemplify this in a multi-contact scenario where several tactile sensors simultaneously touch an object, and we must incorporate all these observations into the object’s pose estimation.”

Essentially, the method devised by this team of researchers can simulate contact information simply based on an object’s 3-D shape. As a result, it does not require any previous tactile data gathered while closely examining the object. This allows the technique to generate pose estimates for an object from the first time it is touched by a robot’s tactile sensors.

“We realized that tactile sensing can be extremely discriminative and produce highly accurate pose estimations,” Bauza said. “While vision will sometimes suffer from occlusions, tactile sensing does not. As a result, if a robot contacts a part of an object that is very unique, i.e., no other touch on the object would look similar to it, then our algorithm can easily identify the contact and thus the object’s pose.”

As many objects have non-unique regions (i.e., the way in which they are positioned can result in very similar contacts), the method developed by Bauza and her colleagues predicts pose distributions, rather than single pose estimates. This particular feature is in stark contrast with previously developed approaches for object pose estimation, which tend to only gather single pose estimates. Moreover, the distributions predicted by the MIT team’s framework can be directly merged with external information to further reduce uncertainty about an object’s pose.

“Notably, we also observed that combining several contacts simultaneously, as it happens when using several fingers to contact an object, rapidly decreases any uncertainty on an objects’ pose,” Bauza said. “This validates our intuition that adding contacts on an object constrains its pose and eases estimation.”

In order to assist humans in their daily activities, robots should be able to complete manipulation tasks with high precision, reliability and accuracy. As manipulating objects directly implies touching them, developing effective techniques to enable tactile sensing in robots is of key importance.

“The ability to sense touch has recently received great interest from industry, and our work achieves this via a combination of three factors: (1) a high-resolution but inexpensive sensing technique based on using small cameras to capture the deformation of a touch surface (e.g., GelSight sensing); (2) recent compact integration of this sensing technique into robot fingers (e.g., GelSlim fingers); (3) and a computational framework based on deep-learning to process effectively the high-resolution tactile images for tactile localization of known parts (e.g., this work),” Alberto Rodriguez, another researcher involved in the study, told TechXplore. “This type of technology is becoming mature and the industry is seeing the value for automating tasks that require precision such as in assembly automation.”

The technique devised by this team of researchers allows robots to estimate the pose of objects they are manipulating in real-time, with high levels of accuracy. This gives a robot the chance to make more accurate predictions about the effects of its movements or actions, which could enhance its performance in manipulation tasks.

To work, the method created by Bauza and her colleagues requires some information about the shape of the object that a robot is manipulating. Therefore, it may prove particularly valuable for implementations in industrial settings, where manufacturers assemble items based on a clear model of their shapes.

In their future work, the researchers plan to extend their framework so that it also incorporates visual information about objects. Ideally, they would like to turn their technique into a visuo-tactile sensing system that can estimate the pose of objects with even greater accuracy.

“Another ongoing work of ours deeply related to this approach aims at exploring the use of tactile perception for complex manipulation tasks,” Bauza said. “In particular, we are learning models that allow a robot to perform accurate pick-and-place operations. The goal is to find object manipulations that not only aim at stable grasps but also aid perception. By using our approach, we can also target grasps that result in discriminative contacts which will improve tactile localization.”

Original article published on TechXplore: https://techxplore.com/news/2021-01-technique-robots-pose.html

For humans, it can be challenging to manipulate thin, flexible objects like ropes, wires, or cables. But if these problems are hard for humans, they are nearly impossible for robots. As a cable slides between the fingers, its shape is constantly changing and the robot’s fingers must be constantly sensing and adjusting the cable’s position and motion.

Standard approaches have used a series of slow and incremental deformations as well as mechanical fixtures to get the job done. Researchers have developed a system that uses a pair of soft robotic grippers with high-resolution tactile sensors (and no added mechanical constraints) to successfully manipulate freely moving cables.

The team first built a two-fingered gripper. The opposing fingers are lightweight and quick moving, allowing nimble, real-time adjustments of force and position. On the tips of the fingers are vision-based GelSight sensors built from soft rubber with embedded cameras. The gripper is mounted on a robot arm, which can move as part of the control system.

The second step was to create a perception-and-control framework to allow cable manipulation. For perception, they used the GelSight sensors to estimate the pose of the cable between the fingers and to measure the frictional forces as the cable slides. Two controllers run in parallel: one modulates grip strength while the other adjusts the gripper pose to keep the cable within the gripper.

When mounted on the arm, the gripper could reliably follow a USB cable starting from a random grasp position. Then, in combination with a second gripper, the robot can move the cable hand-over-hand (as a human would) in order to find the end of the cable. It could also adapt to cables of different materials and thicknesses.

The robot performed an action that humans routinely do when plugging earbuds into a cellphone. Starting with a free-floating earbud cable, the robot was able to slide the cable between its fingers, stop when it felt the plug touch its fingers, adjust the plug’s pose, and finally insert the plug into the jack.

Cable-following is challenging for two reasons. First, it requires controlling the grasp force (to enable smooth sliding) and the grasp pose (to prevent the cable from falling from the gripper’s fingers). This information is hard to capture from conventional vision systems during continuous manipulation because it’s usually occluded, expensive to interpret, and sometimes inaccurate. Also, this information can’t be directly observed with just vision sensors, hence the team’s use of tactile sensors. The gripper’s joints are also flexible, protecting them from potential impact. The algorithms can also be generalized to different cables with various physical properties like material, stiffness, and diameter, and also to those at different speeds.

When comparing different controllers applied to the team’s gripper, their control policy could retain the cable in hand for longer distances than three others; for example, an open-loop controller only followed 36 percent of the total length, the gripper easily lost the cable when it curved, and it needed many re-grasps to finish the task.

The team observed that it was difficult to pull the cable back when it reached the edge of the finger because of the convex surface of the GelSight sensor. They hope to improve the finger-sensor shape to enhance the overall performance. They also plan to study more complex cable manipulation tasks such as cable routing and cable inserting through obstacles and eventually explore autonomous cable manipulation tasks in the automotive industry.

Original article published on Tech Briefs: https://www.techbriefs.com/component/content/article/tb/supplements/md/briefs/37388

The human finger is incredibly sensitive, with a dense array of receptors in the tip that allow us to feel features as small as nanometers. When Kimo Johnson visited aerospace companies that machine metal parts, he learned that many were relying on experienced workers to conduct a “fingernail test” to quickly judge whether a surface defect was just a small scratch or deep enough to potentially compromise the part.

Johnson’s Boston-area startup, Gelsight, aims to provide a more accurate but equally fast alternative with a handheld probe it’s developed. The tip contains a clear gel pad that, like our fleshy fingertips, can conform around an object. Using computer vision techniques, the system “turns touch into an image,” says Johnson, allowing it to map surface features in 3D and measure them down to the single-digit micron level. 

The company is starting to get traction: A number of large aerospace companies are moving toward deploying the tool, Johnson says, including airplane engine maker Rolls-Royce.

Now Gelsight has raised $10 million in a B round that takes its total funding to $11.8 million. It’s expecting revenue to top $4 million this year on sales of over 100 systems and 2,500 gel cartridges.

While the 42-year-old Johnson hopes aerospace will be Gelsight’s breakthrough application, in the nine years since he co-founded the company based on his postdoctoral work at MIT, benchtop versions of its technology have been used for a wide array of tasks, including by Harvard researchers to map fish scales and cosmetics developers to measure how well a product filled in wrinkles.

Rolls-Royce is distributing the handheld device to service reps in its maintenance shops to assess engine parts for damage. Until now, to get an accurate measurement of scratches and dents in sometimes hard to reach places, Rolls and other companies have used a method akin to how dentists make molds of teeth, pressing a wet composite material against the spot that hardens, creating a negative of the defect that is sent off to a lab for cross-sectioning and measurement. It can take as much as 24 hours to get a result.

“We haven’t found another product that has the capability of Gelsight,” says Alistair Donaldson, a Rolls-Royce executive focused on innovation who’s helped the startup hone its technology for practical use. “I could put it in the hands of a child and he could take an accurate measurement.”   

The device is based on work by MIT professor Edward Adelson, a specialist in computer vision who became fascinated by how the sense of touch worked while raising his children. After learning from colleagues that no one yet had developed an artificial tactile sensor with anywhere near the ability of human skin, Adelson set out to do so, leveraging his vision expertise.

“The concept was what would it look like if you had a camera inside your skin and you could see when something’s poking into your skin,” says Johnson, who joined Adelson’s lab as a post-doctoral fellow in 2008 after earning a Ph.D. in computer science from Dartmouth.

Adelson developed a clear pad made of a rubbery thermoplastic elastomer that was coated on one face with a metallic paint. When an object is pressed into the painted side, it takes on the object’s shape and presents a surface that a camera can easily capture a high-fidelity image of through the gel, without distortions from variations of reflective or transparent elements in the object. Johnson built the hardware and wrote algorithms to accurately map surfaces using image processing and machine learning techniques.

Along with a 2011 paper describing the system, Johnson published YouTube videos that caught wider attention, showing renderings of the complex geometry of an Oreo cookie and the raised structure of a printed letter on a $20 bill. He says that led the team to receive thousands of messages from companies and engineers asking if it could be used to measure everything from features of human skin to the quality of abrasives, and a visit by two Boeing engineers with a suitcase full of parts.

“That’s what prompted me to start this company,” Johnson says.    

He launched Gelsight in 2011 with CTO Janos Rohaly, a former lecturer at MIT who cofounded a dental 3D imaging startup that was acquired by 3M, and former CEO Bill Yost. (Johnson has been CEO since 2016.) They made a series of one-off benchtop machines for corporate R&D departments with gel pads optimized for specific applications, such as a toothpaste maker that squished globs of paste to assess its mix of abrasives. But Johnson says he came to realize this approach was a dead end: each company generally only wanted one device that wasn’t widely saleable.

A cellphone manufacturer in China offered a bigger opportunity in mass production, asking Gelsight in 2015 to make a system that could check whether components like metal buttons were the right size. That led Johnson to an epiphany.

“We were making beautiful 3D maps of surfaces, but what the customer often wants is just a single number: What’s the depth of this feature, what’s the chamfer angle,” he says.

They began looking for other industries where a single measurement was needed, day in and day out. That led them to aerospace, particularly scratches and dents on high-value metal parts.

With the proceeds from the sales to the Chinese cellphone maker and lessons learned on what it took to make a system robust and simple enough to be used by ordinary workers, they made the handheld version of the device, debuting it in 2017. 

“Now rather than a single customer we’ve got the entire aerospace market, and that gives us a much better ability to grow the company,” Johnson says. 

Donaldson of Rolls-Royce says he could see hundreds of the $40,000 Gelsight handheld systems being used at his company. Gelsight also earns money on sales of replacement gel cartridges, which wear out after roughly a week of full-time use and go for $120 apiece. 

In conjunction with the $10 million fundraising round, which was led by the venture capital firm Anzu Partners, Gelsight added two former high-ranking military aviation officers to its board to help it crack the defense market, where the pressure is intense to quickly inspect hard-used aircraft and get them back in the air. “If a stone hits the blade of a helicopter, you need to know if this thing can fly,” says Johnson.

With the fresh funding, Johnson aims to double Gelsight’s current headcount of eight and build a smaller version of the wand to get into tighter spaces. He’d like to make it the same size as a human finger. “People use their finger a lot in manufacturing to feel different spaces, if they could have a digital tool that’s a similar form factor, that would be ideal,” he says.

The company is also adapting the technology to make robotic hands that can judge hardness, making them better able to handle delicate and small objects.

There are plenty of other technologies capable of fixed-station inspection, but Gelsight faces little competition from portable alternatives apart from Arizona-based 4D Technologies, which makes a handheld, laser-based device.

It’s been a twisting journey for Johnson. A saxophonist and guitarist, he double-majored in music and math at the University of New Hampshire and had contemplated trying to make a career in music until he developed an interest in computer science while studying for a master’s in electro-acoustic music at Dartmouth. Leaving academia to found a business is another move he wasn’t planning, but he says startup life reminds him of being in a rock band.

“You have a small group of people, you’re trying to break into a space and the odds are stacked against you,” he says. Fortunately, Gelsight has dodged rock band-style interpersonal drama, he says. “We’ve got a great team.”

Original article published in Forbes: https://www.forbes.com/sites/jeremybogaisky/2020/01/23/vision-through-touch-startup-gelsight-hones-measurement-system-that-could-speed-up-aerospace-work/?sh=620f05954ea8