A new model-free framework reorients over 2000 diverse objects with both the...
A new model-free framework reorients over 2000 diverse objects with both the hand facing upward and downward, in a step towards more human-like manipulation.
Source: MIT Computer Science & Artificial Intelligence Lab

Robotic hands manipulate objects with ease

Scientists from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) have created a system that can reorient over two thousand different objects, with the robotic hand facing both upwards and downwards.

At just one year old, a baby is more dexterous than a robot. Sure, machines can do more than just pick up and put down objects, but we're not quite there as far as replicating a natural pull towards exploratory or sophisticated dexterous manipulation goes.

OpenAI gave it a try with "Dactyl" (meaning "finger" from the Greek word daktylos), using their humanoid robot hand to solve a Rubik's cube with software that's a step towards more general AI, and a step away from the common single-task mentality. DeepMind created "RGB-Stacking," a vision-based system that challenges a robot to learn how to grab items and stack them.

Scientists from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) created a framework that's more scaled up: a system that can reorient over two thousand different objects, with the robotic hand facing both upwards and downwards. This ability to manipulate anything from a cup to a tuna can, and a Cheez-It box, could help the hand quickly pick-and-place objects in specific ways and locations—and even generalize to unseen objects.

This deft "handiwork"—which is usually limited by single tasks and upright positions—could be an asset in speeding up logistics and manufacturing, helping with common demands such as packing objects into slots for kitting, or dexterously manipulating a wider range of tools. The team used a simulated, anthropomorphic hand with 24 degrees of freedom, and showed evidence that the system could be transferred to a real robotic system in the future.

"In industry, a parallel-jaw gripper is most commonly used, partially due to its simplicity in control, but it's physically unable to handle many tools we see in daily life," says MIT CSAIL Ph.D. student Tao Chen, member of the Improbable AI Lab and the lead researcher on the project. "Even using a plier is difficult because it can't dexterously move one handle back and forth. Our system will allow a multi-fingered hand to dexterously manipulate such tools, which opens up a new area for robotics applications."

Give me a hand

This type of "in-hand" object reorientation has been a challenging problem in robotics, due to the large number of motors to be controlled and the frequent change in contact state between the fingers and the objects. And with over two thousand objects, the model had a lot to learn.

The problem becomes even more tricky when the hand is facing downwards. Not only does the robot need to manipulate the object, but also circumvent gravity so it doesn't fall down.

The team found that a simple approach could solve complex problems. They used a model-free reinforcement learning algorithm (meaning the system has to figure out value functions from interactions with the environment) with deep learning, and something called a "teacher-student" training method.

For this to work, the "teacher" network is trained on information about the object and robot that's easily available in simulation, but not in the real world, such as the location of fingertips or object velocity. To ensure that the robots can work outside of the simulation, the knowledge of the "teacher" is distilled into observations that can be acquired in the real world, such as depth images captured by cameras, object pose, and the robot's joint positions. They also used a "gravity curriculum," where the robot first learns the skill in a zero-gravity environment, and then slowly adapts the controller to the normal gravity condition, which, when taking things at this pace—really improved the overall performance.

"Our results show that one can learn robust control strategies that are shape agnostic."

Pulkit Agrawal

While seemingly counterintuitive, a single controller (known as brain of the robot) could reorient a large number of objects it had never seen before, and with no knowledge of shape.

"We initially thought that visual perception algorithms for inferring shape while the robot manipulates the object was going to be the primary challenge," says MIT professor Pulkit Agrawal, an author on the paper about the research. "To the contrary, our results show that one can learn robust control strategies that are shape agnostic. This suggests that visual perception may be far less important for manipulation than what we are used to thinking, and simpler perceptual processing strategies might suffice."

Many small, circular shaped objects (apples, tennis balls, marbles), had close to one hundred percent success rates when reoriented with the hand facing up and down, with the lowest success rates, unsurprisingly, for more complex objects, like a spoon, a screwdriver, or scissors, being closer to thirty.

Beyond bringing the system out into the wild, since success rates varied with object shape, in the future, the team notes that training the model based on object shapes could improve performance.

Subscribe to our newsletter

Related articles

ReSkin helps to discover a sense of touch

ReSkin helps to discover a sense of touch

Carnegie Mellon University and Meta AI (formerly Facebook AI) want to increase the sense of touch in robotics, wearables, smart clothing and AI.

Teaching robots to think like us

Teaching robots to think like us

Advances in physical reservoir computing could contribute to creating artificial intelligence machines that think like us.

How do people emotionally respond to cloned faces?

How do people emotionally respond to cloned faces?

Researchers examined people’s emotional response to cloned faces, which could soon become the norm in robotics.

Expanding human-robot collaboration in manufacturing

Expanding human-robot collaboration in manufacturing

To enhance human-robot collaboration, researchers at Loughborough University have trained an AI to detect human intention.

Bipedal robot learns to run

Bipedal robot learns to run

Cassie the robot has made history by traversing 5 kilometers, completing the route in just over 53 minutes.

A contact aware robot design

A contact aware robot design

Researchers have developed a new method to computationally optimize the shape and control of a robotic manipulator for a specific task.

Companion robot to support primary care providers

Companion robot to support primary care providers

Intuition Robotics announced a significant expansion of ElliQ, their AI-driven companion robot, to enable the extension of primary care teams' presence into older adults' homes.

A robot scientist is ready for drug discovery

A robot scientist is ready for drug discovery

The robot scientist Eve has been assembled and is now operating at Chalmers University of Technology. Eve’s f​irst mission is to identify and test drugs against Covid-19.​

Engineers build smart robotic exoskeletons

Engineers build smart robotic exoskeletons

Researchers are developing exoskeletons and prosthetic legs capable of thinking and making control decisions on their own using AI technology.

Popular articles

Subscribe to Newsletter