Vibration Suppression in Collaborative Flexible Payload Manipulation Using Passive Force Control

Imagine you are trying to move a giant, wobbly jellyfish across a room. If you grab it and yank it quickly, it doesn't just move; it sloshes, swings, and vibrates wildly. In the world of robotics, this is a huge problem. When robots try to carry heavy, flexible things—like giant metal sheets or parts of a nuclear reactor—they can't just be strong; they have to be gentle. If they move too jerkily, the object starts shaking like a gelatin dessert, which can damage the object or make the robot lose its grip. This is especially tricky when the object is so big and heavy that it bends under its own weight, and the space to move it is incredibly tight, leaving no room for error. Scientists have been trying to figure out how to move these "wobbly giants" without making them shake, but most solutions either require expensive sensors that can't survive in harsh environments or only work in computer simulations.

This paper tackles that exact problem by teaching two different robots how to work together as a team to carry a flexible load without the wobble. The researchers propose a clever "leader-follower" strategy. Think of it like a dance where a strong, heavy-duty robot (the leader) does the heavy lifting and decides where to go, while a lighter, more sensitive robot (the follower) acts like a shock absorber. Instead of the follower fighting the leader, it uses a special control system called "admittance control" to go with the flow, gently damping out the vibrations as they happen. To make sure the leader doesn't accidentally start the shaking in the first place, the team also uses a technique called "input shaping," which is like carefully timing the steps of the dance so the music never hits a note that makes the jellyfish jiggle. By combining these two tricks, the robots can move massive, floppy objects smoothly, even if they can't see the vibrations directly.

The Team-Up Strategy

The core idea of this research is to use two very different robots to move a single, flexible object. In the real world, moving something like a 10-meter-long, 80-tonne "breeding blanket" (a critical part of a future fusion energy reactor) is a nightmare. These blankets are so big they bend, and after years of use, they lose their stiffness, becoming even more like giant, floppy noodles. The challenge is that you can't just stick sensors all over them to measure the shaking because the environment is full of dangerous radiation that would fry the electronics.

So, the authors designed a system where a "heavy-duty" robot acts as the leader. It grabs one end of the flexible object and pulls it along a path. A second, lighter robot acts as the follower, holding the other end. Instead of the follower trying to guess where to go, it listens to the leader. The leader feels the force it's exerting on the object, and the follower uses a special "admittance controller" to react to that force. It's like if you were walking with a friend holding a long, heavy pole between you. If your friend stumbles, you don't pull back hard; you adjust your step to keep the pole steady. Here, the follower robot is programmed to be "compliant," meaning it yields to the forces it feels, effectively acting as a giant, robotic shock absorber that soaks up the energy of the vibrations.

The Two-Pronged Attack on Wobble

The paper suggests that using just one trick isn't enough, so they combine two methods to kill the vibrations.

First, they tackle the vibrations after they start using the follower robot. The follower doesn't just follow blindly; it uses a "force estimator" to guess how much force is being applied based on the tiny gap between where the leader is and where the follower is. It then moves to close that gap, but it does so with a specific amount of "damping." Imagine pushing a swing: if you push it at the wrong time, it goes crazy. If you push it with just the right resistance, it slows down. The follower robot is tuned to be that resistance, turning the kinetic energy of the shaking into heat (dissipation) so the object stops wiggling faster. The researchers proved mathematically that this system is "passive," meaning it naturally loses energy over time rather than gaining it, ensuring the vibrations die out on their own.

Second, they tackle the vibrations before they start by being careful about how the leader moves. Even if the follower is great at damping, a bad start can make the object shake violently. To prevent this, the leader uses "input shaping." This is a bit like a drummer who knows exactly when to hit the drum to avoid a bad rhythm. The computer takes the command to move the robot and splits it into two parts: a small push, a short pause, and then the rest of the move. This timing is calculated based on how long it takes the object to vibrate once (its natural period). By delaying the second part of the move by exactly half a vibration cycle, the second push cancels out the wobble created by the first push. It's a "feed-forward" trick, meaning it plans the move perfectly in advance without needing to wait for sensors to tell it something is wrong.

What the Experiments Showed

The team tested this idea in two ways: first in a computer simulation using Matlab/Simulink, and then in a real lab with actual robots. They used a KUKA Quantek robot as the heavy leader and a KUKA-Iiwa robot as the light follower, trying to move a flexible acrylic sheet. The goal was to move the sheet 10 centimeters in just 0.5 seconds—a very fast, jerky move that usually causes a lot of shaking.

They compared four different ways of controlling the robots: doing nothing, using only the follower's damping (admittance), using only the leader's careful timing (input shaping), and using both together.

The results were clear. When they did nothing, the sheet shook with a peak amplitude of about 28.8 mm in the simulation and 26.3 mm in the real experiment. When they used just the follower's damping, the shaking dropped significantly (to 15.6 mm in simulation and 11.4 mm in the lab). When they used just the leader's careful timing, it also helped, though in the real world, it was less effective at stopping the initial peak shake than the follower alone. However, the real winner was the combination. When they used both the follower's damping and the leader's careful timing together, the vibrations were crushed. In the simulation, the peak shake was reduced by nearly 80% (down to 5.8 mm), and in the real experiment, it was reduced by a similar 79.5% (down to 5.4 mm). The time it took for the shaking to stop completely (settling time) also dropped dramatically, from over 4 seconds with no control to just over 1 second with the combined method.

Why This Matters

The paper concludes that this "leader-follower" approach, which combines a smart follower that acts as a shock absorber with a leader that plans its moves carefully, is a highly effective way to move heavy, flexible objects. It's particularly useful because it doesn't rely on fragile sensors placed directly on the object, which is a huge advantage for dangerous environments like nuclear reactors. While the current tests were done with a flat sheet and 2D movement, the authors suggest this method could be scaled up to handle the massive, 3D structures needed for future fusion power plants. They plan to test it with a 3D-printed model of a reactor blanket next, hoping to prove that their "robot dance" can keep even the wobbliest giants steady.