IRIS — Simulation

Quick Start

Run the simulation.

Python ≥ 3.9, MuJoCo ≥ 2.3.

# Install
cd mujoco_sim && pip install -r requirements.txt

# Cinema motion modes
python cinema_line_tracking.py --mode crane
python cinema_line_tracking.py --mode dolly
python cinema_line_tracking.py --mode pan

# Circular path tracking
python circle_path_tracking.py

# Teleoperation
python teleop_ik.py    # Cartesian IK
python teleop_fk.py    # Joint-space

Kinematics

FK & Numerical IK

DH parameters shared across the MuJoCo XML, ROS TF tree, and kinematics solvers. FK is closed-form from DH; IK is iterative numerical (Damped Least-Squares Jacobian via MuJoCo's mj_jacBody).

DH Parameter Table

Joint	Description	aᵢ (m)	αᵢ (°)	dᵢ (m)	θ_off (°)
J1	Base yaw	0.0000	0.0	0.2487	0.0
J2	Shoulder pitch	0.0218	90.0	0.0590	180.0
J3	Arm pitch	0.2998	0.0	0.0000	0.0
J4	Elbow pitch	0.0200	90.0	0.0000	0.0
J5	Wrist pitch	0.3251	−90.0	0.0000	0.0
J6	Wrist roll	0.0428	90.0	0.0000	0.0

Max reach ~1.0 m · Continuous base yaw

FK / IK

# Forward kinematics — closed-form DH
python kinematics/forward_kinematics.py \
  --q 0 0 0 0 0 0

# Inverse kinematics — numerical DLS Jacobian
python inverse_kinematics_numerical.py \
  --target_xyz 0.6 0.0 0.5

Numerical IK: iterative Damped Least-Squares (DLS) Jacobian. Singularity-robust via damping constant λ=10⁻⁴. Converges in <100 iterations to 0.1 mm tolerance.

Joint-space teleoperation (FK)

Cartesian teleoperation (IK)

Cinema Motion Modes

Built for filmmakers.

Three cinematography-specific motion modes. Each generates smooth camera paths used in professional production.

Crane

Vertical arc sweeping the camera upward while tracking a fixed point. Used for reveals and establishing shots.

python cinema_line_tracking.py --mode crane

Dolly

Linear push-in / pull-out while maintaining camera orientation. Simulates a physical dolly track.

python cinema_line_tracking.py --mode dolly

Pan

Lateral arc sweep holding a subject in frame. Smooth rotation for product shots and subject tracking.

python cinema_line_tracking.py --mode pan

Circular Path

End-effector follows a circular arc around a target point. Configurable radius and center.

python circle_path_tracking.py \
  --radius 0.15 \
  --center 0.45 0.0 0.4

Path Planning

Motion Planners

RRT* — Optimal Rapidly-Exploring Random Tree

Asymptotically optimal sampling-based planner. Finds collision-free paths in joint space with a rewiring step that continually improves path cost.

cd classical_planner
python rrt.py

Artificial Potential Field

Gradient descent on an artificial potential field with attractive goal forces and repulsive obstacle forces. Fast reactive replanning for quasi-static environments.

python path_tracking.py