Text settings Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only Learn more Minimize to nav Switching from one smartphone to another is mostly a smooth procedure. You log into your accounts and your apps, preferences, and contacts should sync to the new hardware. But in the world of robotics, swapping an old robotic arm for a newer model has meant setting everything up from scratch.
To fix that, a team of researchers at the Swiss École Polytechnique Fédérale de Lausanne (EPFL) has developed what they call Kinematic Intelligence, a framework that makes switching robots work more like switching smartphones. They describe their system in a recent Science Robotics paper.
For years, roboticists have been working on getting robots to learn from demonstration—teaching them new skills by showing them what to do, rather than writing lines of code. The idea is to remotely control or physically guide the robot’s arm to teach it a task like wiping a table, stacking boxes, or welding a car component. The problem is that most of these taught skills end up tied to the specific robot the training was done with.
But robotics is advancing quickly. “The robots have different designs, and nowadays there are new designs being proposed—that brings its own set of challenges,” said Sthithpragya Gupta, a roboticist at EPFL and lead author of the study. If a new robot has slightly longer links, a different joint orientation, or a more complex configuration, that learned behavior instantly breaks and the new robot will likely flail, freeze, or crash if attempting it.
“With new designs come different capabilities and constraints,” said Durgesh Haribhau Salunkhe, an EPFL roboticist and co-author of the study. “The problem is to adapt to these constraints and capabilities—to faithfully replicate the actions demonstrated by a human.” Today, making the leap from one robot body to another usually means starting from scratch and retraining the whole system.
When a robot moves through space to complete a task, it must constantly calculate how to bend its joints to keep its end-effector (a robotic equivalent of a hand) on the right path. The robot has to avoid hitting a physical limit, or worse, a singularity, which in robotics is a mathematical danger zone: a physical configuration where the robot’s joints align in such a way that it temporarily loses a degree of freedom. “In such positions, the robot’s motion may become unstable or [you] may lose control of the robot,” Gupta said.
In human terms, it works roughly like locking the elbows as they get fully straightened when pushing something heavy, which makes the arms unable to perform side-to-side movements for a moment.
Transferring skills from one robot to another is hard because differently structured robots usually have a different topology of singularities. When a robot’s algorithm blindly follows a path and hits a singularity, the math controlling its joints will fail. The robot might try to spin a joint at infinite speed, for instance, resulting in a sudden, unsafe movement. Gupta’s team solved this by giving the robots a deep, innate mathematical awareness of their own physical limitations. This Kinematic Intelligence, as they call it, lets a user demonstrate a skill just once, and have it executed safely by an entirely different type of robot.
And (surprisingly, these days) Kinematic Intelligence was built in an AI-free manner.
Traditionally, engineers have dealt with singularities through software fixes. They built inverse models, complex mathematical formulas that work backward from the target position of the robot’s end-effector to map all the joint positions required to get it there. Then, they just slapped on safety filters or corrections to prevent the robot from getting itself into trouble.
Some of the newer, data-driven AI approaches take less effort and expertise but require access to every robot that the control software will be used on during the training phase. “Also, there is this probabilistic or black box nature of AI wherein it can do something incoherent, which can be potentially catastrophic,” Gupta said. His team wanted certainty, not probabilities, so they took a different approach.
Instead of trying to correct for a robot’s mechanical constraints after the training, they embedded these constraints directly into the control policy from the beginning. They focused on three-revolute robots—basically robotic arms with three joints—which act as the foundational building blocks for many of the commercial robots we see today. Through an algebraic analysis of the robots’ parameters, such as the lengths of their links and the offsets of their joints, the team mapped out exactly where the singularities lie within their joint space. These singularities, combined with the hard limits of the joints, slice the robot’s possible movement space into feasible regions the researchers call aspects.
By looking at the topology of these aspects, the researchers classified three-revolute robots (those with three joints) into six categories. This way, once they knew which of these six categories a specific robot falls into, they instantly knew the exact structure of its physical limitations—a complete map of its danger zones.
Armed with this map, the Kinematic Intelligence framework enables robots to go around their singularities using a strategy the team calls a track cycle. Based on its category classification, the robot knows its physical limits, which prevents it from crashing and dynamically redirects the movement to safely slide or traverse along the edge of the singularity boundary. The robot carefully follows this boundary until it finds a safe configuration where it can re-enter the nominal path to finish the task.
When the team made sure the math behind their idea was correct, they put their framework to the test on various machines. And it worked.
The experimental setup included a compact 6-DoF Duatic DynaArm with tight joint limits, a 7-DoF KUKA LWR IIWA 7 with moderate limits, and a 7-DoF Neura Robotics Maira M with much more relaxed boundaries. With these machines, the researchers built a mock multi-robot assembly line where three different robotic arms cooperated to complete a sequence of tasks. At the beginning, a human performed a single demonstration of three skills in sequence. “We demonstrated a task where you push something off a conveyor belt, pick it up and put it on a workbench, and then pick it up again and throw it into a basket,” Gupta said. All these actions were then distributed among the robots so that each robot performed one of them: the DynaArm did the pushing, the KUKA did the picking and placing, and the Neura did the picking and throwing.
Even though the pushing and throwing motions forced the robots into excursions near the boundaries of their physical workspaces, and the pick-and-place maneuver demanded complex internal mathematical checks, all three machines were able to learn a functional policy via a single human demonstration. “And then we said, you know what, let’s shuffle these robots around,” Gupta said.
Without any retraining, the team swapped the robots’ locations and tasks. It turned out their Kinematic Intelligence made it possible to complete the sequence when KUKA was responsible for pushing, the DynaArm for throwing, the Neura for picking and placing, and in all other possible configurations. “The key challenge for now is to take this technology to the industrial assembly floor,” Gupta said. He admitted, though, that there are several details the team still has to figure out.
While the Kinematic Intelligence framework guarantees mechanically safe motion, it currently lacks the advanced sensing and context-sensitive decision-making required for unpredictable environments. While the researchers acknowledge that the system flawlessly handles a robot’s internal physical constraints like singularities and joint limits, it is not yet equipped to inherently understand the nuances of the objects it interacts with. For example, the system cannot currently distinguish between moving a full container, which requires slow, careful handling, and an empty one, which can be moved quickly. What’s more, it requires the integration of high-level cognitive safety checks to integrate human commands with common sense, such as knowing not to grab a knife when asked to prepare coffee.
Another hurdle to overcome before Kinematic Intelligence can transition from controlled laboratory experiments to factory floors is the integration of advanced environmental sensing, which would enable robots to safely navigate dynamic spaces where humans are constantly and unpredictably moving around. Additionally, while the software framework has already been validated on current industrial robots, its deployment in more sensitive fields like medicine is currently bottlenecked by hardware limitations.
“If we talk deploying this technology in medical scenarios, I believe in the next five years we will see mechanically safer robots that should make this possible,” Salunkhe said. “Our framework can be immediately translated to such new designs, so we’re waiting for these robots now.”
Gupta’s and Salunkhe’s work on robots’ skill sharing is published in Science Robotics: http://dx.doi.org/10.1126/scirobotics.aea1995
---
**İlgili Kaynaklar:**
yapay zeka danışmanlık ve çözüm hizmetleri konusunda [yapay zeka firması](https://yapayzekafirmasi.com) ile iletişime geçebilirsiniz.
