udaan.models.mujoco package

Submodules

• udaan.models.mujoco.multi_quad_cs_pointmass module
  • MultiQuadrotorCSPointmass
• udaan.models.mujoco.multi_quad_rigidbody module
  • MultiQuadRigidbody
• udaan.models.mujoco.quadrotor_comparison module
  • QuadrotorComparison
• udaan.models.mujoco.quadrotor_cspayload_fleet module
  • QuadrotorCsPayloadFleet
• udaan.models.mujoco.quadrotor_fleet module
  • QuadrotorFleet

Module contents

MuJoCo physics backend for Udaan.

Requires mujoco package. Install with: pip install udaan

class udaan.models.mujoco.MujocoModel

Bases: object

__init__(model_path, render=False, record=None)

add_arrow_at(p, R, s, label='', color=None)

Visual markers not yet supported with built-in viewer.

add_marker_at(p, size=None, rgba=None, label='')

Visual markers not yet supported with built-in viewer.

reset()

set_overlay(text)

Set overlay text displayed in the top-left corner.

wait_for_close()

Keep the viewer open until the user closes the window. If recording, save and close immediately instead of holding.

class udaan.models.mujoco.MultiQuadRigidbody

Bases: BaseModel

Multi-quadrotor rigid-body payload MuJoCo model.

4 quadrotors carrying a rigid-body payload connected via cables. Outer loop: position-level thrust vectors (3 per quad). Inner loop: geometric attitude controller converts to wrench.

class State

Bases: object

class CableState

Bases: object

__init__(**kwargs)

reset()

class Odometry

Bases: object

__init__(**kwargs)

reset()

__init__(nQ, **kwargs)

reset()

__init__(**kwargs)

compute_attitude_control(i, thrust_force)

Geometric attitude controller for quadrotor i.

Parameters:

• i – quadrotor index

• thrust_force – (3,) desired thrust vector in world frame

Returns:

(4,) [thrust_scalar, Mx, My, Mz] wrench

quad_position_control()

Basic PD hover controller for each quadrotor.

reset(**kwargs)

simulate(tf, **kwargs)

step(action)

Step with position-level thrust vectors (3*nQ).

Parameters:

action – (3*nQ,) thrust force vector per quadrotor.

class udaan.models.mujoco.MultiQuadrotorCSPointmass

Bases: BaseModel

Multi-quadrotor with cable-suspended pointmass payload (MuJoCo).

class State

Bases: object

class Cable

Bases: object

__init__(**kwargs)

reset()

class Quadrotor

Bases: object

__init__(**kwargs)

reset()

__init__(nQ, **kwargs)

reset()

__init__(**kwargs)

compute_attitude_control(i, thrust_force)

quad_position_control()

quadrotor position control

reset(**kwargs)

reset state and time

simulate(tf, **kwargs)

step(u)

udaan.models.mujoco.Quadrotor

alias of QuadrotorMujoco

class udaan.models.mujoco.QuadrotorComparison

Bases: BaseModel

__init__(**kwargs)

add_reference_marker(x, label='xQd')

reset(**kwargs)

set_inertia(I, plant=True)

set_mass(m, plant=True)

step(u)

class udaan.models.mujoco.QuadrotorCsPayloadFleet

Bases: BaseModel

Multiple independent quadrotor-cable-payload agents in one MuJoCo scene.

Each agent keeps its own QuadrotorCsPayloadBase for controller/state bookkeeping; the fleet owns a single MuJoCo model containing N quads + N cables + N payloads.

Usage:

fleet = QuadrotorCsPayloadFleet(num_agents=2, render=True)
fleet[0]._payload_controller.setpoint = lambda t: (target, 0, 0)
fleet[1]._payload_controller.setpoint = lambda t: (target, 0, 0)
fleet.simulate(tf=8.0, payload_positions=[[1,1,0.5], [-1,-1,0.5]])

__init__(**kwargs)

reset(**kwargs)

simulate(tf, **kwargs)

Run closed-loop sim. Each agent uses its own _payload_controller.

step(u)

u is (4 * nQ,): stacked [thrust, Mx, My, Mz] per agent.

class udaan.models.mujoco.QuadrotorFleet

Bases: BaseModel

Multiple independent quadrotors in a single MuJoCo scene.

Each quadrotor has its own controllers and can optionally carry an added mass at an offset (for robustness testing).

Usage:

fleet = QuadrotorFleet(num_quadrotors=3, render=True)
fleet[0].position_controller = MyController(...)

# Add disturbance to quad 1: 0.25kg at [0.2, 0.2, -0.05]
fleet = QuadrotorFleet(
    num_quadrotors=3,
    render=True,
    disturbances={
        1: {"mass": 0.25, "offset": [0.2, 0.2, -0.05]},
    },
)
fleet.simulate(tf=10, position=np.array([0, 0, 1]))

__init__(**kwargs)

reset(**kwargs)

simulate(tf, **kwargs)

Run simulation with each quadrotor using its own controllers.

step(u)

Step all quadrotors. u is (4*nQ,) wrench vector.