Robotics Weekly

Drones tie knots in midair using only cables

No gripper, no arm, just coordinated flight paths and friction.

Aerial manipulation usually means bolting a gripper or an arm onto a drone. A new paper skips both. Several drones each hold one end of a cable and fly coordinated paths that wrap and interlace it around an object in layers, forming what the authors call a multilayer hitch, posted by Lukas Ziegler.

Lukas Ziegler
@lukas_m_ziegler
X
Friction grows exponentially with each wrap, so a handful of interlaced layers turns a loose loop into a grip strong enough to lift the object.
Nov 15, 2026 · View on X
Lukas Ziegler
@lukas_m_ziegler
X
Aerial manipulation almost always means bolting a gripper or an arm onto a drone.
Nov 15, 2026 · View on X

The cable is the gripper. That is the whole idea.

Why wrapping works

Friction grows exponentially with each wrap. So a handful of interlaced layers turns a loose loop into a grip strong enough to lift the object. The authors back this with a capstan-based model, which gives a conservative rule for how many layers you need before you try to pick something up. Conservative is the right word there. You want the rule to tell you to add one more wrap than strictly necessary, not one fewer.

On top of the hitch, the paper introduces two new actions, autonomous multilayer tying and autonomous release. Put those together and you get a full pipeline. Tie, lift, carry, let go, with no human in the loop. The release half matters as much as the tying half, because a knot you cannot undo is a drone you cannot reuse.

It scales sideways

The formation algorithm runs in parallel, so execution time stays the same whether you use a few drones or many. That is an unusual property for a multi-robot coordination problem, where adding agents often means adding wait time. Here more drones means more wraps in the same window rather than a longer job.

Shown versus claimed

Ziegler says the team showed the whole thing end to end, in simulation and on real hardware, tying, lifting and releasing with cables alone. So this is not a simulation-only result. The post does not give success rates, object weights, drone counts or flight times, so how reliable it is outside the demo is not something the source answers.

The appeal is mostly about what you do not have to carry. A gripper or arm is mass on the airframe, and mass on a drone is flight time you do not get back. A cable is light and it does not care much about the shape of the thing it is wrapping. The trade is that you now need several aircraft flying tight coordinated paths around a shared object, which is its own hard problem, and the paper's answer to that is the parallel formation algorithm.

The paper is linked in Ziegler's post.

Sound like you were at the demo

One email a day. Under three minutes. The demos, the deployments and the part the press release left out.

Get the next one by email

Robotics every day from the people building it. The demos, the deployments and the arguments worth your time. Every claim links back to the engineer.