SimPLE, an approach to object manipulation developed by MIT Department of Mechanical Engineering researchers, aims to “reduce the burden of introducing new objects to make it so that robots can still interact precisely but more flexibly,” said doctoral student Antonia Delores Bronars. (Image: John Freidah/MIT Department of Mechanical Engineering)

Pick-and-place machines are a type of automated equipment used to place objects into structured, organized locations. These machines are used for a variety of applications — from electronics assembly to packaging, bin picking, and even inspection — but many current pick-and-place solutions are limited. Current solutions lack “precise generalization,” or the ability to solve many tasks without compromising on accuracy.

“In industry, you often see that [manufacturers] end up with solutions very tailored to the particular problem that they have, so a lot of engineering and not so much flexibility in terms of the solution,” said Maria Bauza Villalonga, a senior research scientist at Google DeepMind, where she works on robotics and robotic manipulation. “SimPLE solves this problem and provides a solution to pick-and-place that is flexible and still provides the needed precision.”

A new paper by Department of Mechanical Engineering researchers published in the journal Science Robotics explores more precise pick-and-place solutions. In precise pick-and-place, also known as kitting, the robot transforms an unstructured arrangement of objects into an organized arrangement. With this approach, dubbed SimPLE (Simulation to Pick Localize and placE), learns to pick, regrasp and place objects using the object’s computer-aided design (CAD) model, and all without any prior experience or encounters with the specific objects.

“The promise of SimPLE is that we can solve many different tasks with the same hardware and software using simulation to learn models that adapt to each specific task,” said Alberto Rodriguez, an MIT visiting scientist.

Using a dual-arm robot equipped with visuotactile sensing, the SimPLE solution employs three main components: task-aware grasping, perception by sight and touch (visuotactile perception), and regrasp planning. Real observations are matched against a set of simulated observations through supervised learning so that a distribution of likely object poses can be estimated, and placement accomplished.

In experiments, SimPLE successfully demonstrated the ability to pick and place diverse objects spanning a wide range of shapes, achieving successful placements over 90 percent of the time for six objects, and over 80 percent of the time for eleven objects.

“There’s an intuitive understanding in the robotics community that vision and touch are both useful, but [until now] there haven’t been many systematic demonstrations of how it can be useful for complex robotics tasks,” said mechanical engineering doctoral student Antonia Delores Bronars.

“Most work on grasping ignores the downstream tasks,” said Matt Mason, chief scientist at Berkshire Grey and professor emeritus at Carnegie Mellon University who was not involved in the work. “This paper goes beyond the desire to mimic humans and shows from a strictly functional viewpoint the utility of combining tactile sensing and vision with two hands.”

Ken Goldberg, the William S. Floyd Jr. Distinguished Chair in Engineering at the University of California at Berkeley, who was also not involved in the study, said the robot manipulation methodology described in the paper offers a valuable alternative to the trend toward AI and machine learning methods. “The authors combine well-founded geometric algorithms that can reliably achieve high precision for a specific set of object shapes and demonstrate that this combination can significantly improve performance over AI methods,” said Goldberg. “This can be immediately useful in industry and is an excellent example of what I call 'good old-fashioned engineering' (GOFE).”

Bauza and Bronars said this work was informed by several generations of collaboration.

“In order to really demonstrate how vision and touch can be useful together, it’s necessary to build a full robotic system, which is something that’s very difficult to do as one person over a short horizon of time,” said Bronars. “Collaboration, with each other and with Nikhil [Chavan-Dafle Ph.D.] and Yifan [Hou Ph.D.], and across many generations and labs really allowed us to build an end-to-end system.”

Source 

Topics:
Automation Electronics Electronics & Computers Industrial Controls & Automation Machinery Motion Control Positioning Equipment Robotics

More From SAE Media Group

    Transcript

    00:00:00 [MUSIC PLAYING] ANTONIA DELORES BRONARS: Manufacturing is one of the first places where robotics has become very important and very impactful. The general paradigm is one robot is capable of doing one job, interacting with one type of object, which makes it very expensive for companies to introduce new parts to their manufacturing lines.

    00:00:23 Our solution is to reduce the burden of introducing new objects to make it so that robots can interact still precisely, but more flexibly. The approach that we have developed is called SimPLE. And it relies on interacting with objects in simulation to precisely pick up, localize, and place objects. We rely on the object CAD models for the type of domains that we're targeting. It's a compelling demonstration of the power of integrating vision and tactile.

    00:00:54 One gives you a very global, high level view of what's going on, while the other gives you a local but highly precise view of where the object is. One of the features of our system is that it's bimanual, meaning it has two arms. So, if necessary, it can decide to hand the object off to the other arm, which can lead to a higher success rate. Our lab recognized that there is a gap when it comes from transforming arrangements of objects from an unstructured set to a structured set, which is super

    00:01:24 valuable in the industry because we open the door for a wider range of downstream tasks. This system could see applications in manufacturing, in hospitals, in laboratory settings, anywhere where the set of objects that the robot would interact with are relatively consistent over some horizon of time. I think the really cool thing about this work is it truly is an end to end system that puts together a lot of pieces from perception with multiple modalities

    00:01:50 to planning to really explore the synergies between these different parts and how we can leverage the robot's knowledge of how well they work to come up with a plan that is robust and efficient. It's capable of understanding its own capabilities as well as limitations, which is a very humanlike form of intelligence. [MUSIC PLAYING]