A Python MuJoCo-based models and controllers for quadcopter cable-suspended payload systems.

---

---

Developed as part of the thesis: Dynamics and Control for Collaborative Aerial Manipulation (Kotaru, 2022). Original research code: vkotaru/floating_models. This package is the cleaned-up public release, refactored with Claude.

🚁 Quadrotor dynamics  •  🎛 Geometric control on SE(3)/SO(3)  •  ⛓ Cable-suspended payloads  •  🤝 Multi-quad cooperative transport  •  🕹 MuJoCo simulation  •  🖥 CLI interface

Installation

pip install udaan

MuJoCo is included as a core dependency. Install all extras (dev, docs, RL):

pip install udaan[all]

For development:

git clone https://github.com/vkotaru/udaan.git
cd udaan
pip install -e ".[all]"

Quick Start

CLI

# Quadrotor with geometric SE(3) control
udaan run quadrotor                                    # MuJoCo (default)
udaan run quadrotor -m base                            # pure dynamics (no viz)
udaan run quadrotor -m vfx                             # VPython visualization

# Trajectory tracking
udaan run quadrotor --traj hover -p 1,1,0              # hover (default)
udaan run quadrotor --traj spiral -p 0,0,2             # helical spiral
udaan run quadrotor --traj lissajous -p 0,0,2          # 3D Lissajous
udaan run quadrotor --traj circle -p 0,0,1             # circular

# Cable-suspended payload
udaan run quad-payload -t 10 -c tendon                 # spatial tendon cable
udaan run quad-payload -t 10 -c links                  # rigid link chain
udaan run quad-payload -t 10 -c cable                  # composite cable (experimental)
udaan run quad-payload -t 10 -m vfx                   # vpython backend

# Multi-quadrotor cooperative transport
udaan run multi-quad -n 3 -t 10                        # N-quad pointmass payload
udaan run multi-quad-rigid -t 10                       # rigid-body payload

# Fleet: compare controllers side-by-side
udaan run fleet --demo l1-comparison                   # L1 adaptive vs PD
udaan run fleet --demo gain-sweep                      # PD gain comparison
udaan run fleet -n 4 --trail                           # 4 quads with trails

# Cable-payload fleet: compare payload controllers / gains side-by-side
udaan run cspayload-fleet --demo same-gains            # 2 agents, default gains
udaan run cspayload-fleet --demo gain-sweep            # cable kp/kd × [0.5, 1, 1.5, 2]
udaan run cspayload-fleet --demo gain-sweep --same-start  # overlap start, gain-driven divergence
udaan run cspayload-fleet -n 4                         # 4 agents, default gains

# Recording
udaan run quadrotor -t 5 -r out.gif                    # save to GIF
udaan run quadrotor --traj spiral -r spiral.mp4        # save to MP4

Python

from udaan.models.quadrotor import QuadrotorBase, QuadrotorMujoco

# Pure dynamics (no rendering)
mdl = QuadrotorBase()
mdl.simulate(tf=10, position=[1., 1., 0.])

# MuJoCo with visualization
mdl = QuadrotorMujoco(render=True)
mdl.simulate(tf=10, position=[1., 1., 0.])

Documentation

Full API reference, controller roadmap, and tutorials coming soon at udaan.readthedocs.io.

License

BSD 3-Clause License. See LICENSE for details.