Exploration of Franka Emika low-level controllers

Controllers Implemented

1. Joint Velocity Controller

This controller commands angular velocities (rad/s) to each joint using the VelocityJointInterface. It's low-level and typically runs at ~1 kHz, enabling very smooth real-time motion.

• Demo Behavior: All joints commanded the same velocity (±0.1 rad/s), direction flipped every 8 seconds.

• Pros: Simple to implement, very responsive, great for real-time motion generation or teach-by-demonstration.

• Cons: No position feedback — can drift over time.

• Applications: Trajectory tracking, low-level admittance control, real-time input.

2. Joint Position Controller

This controller sends absolute angular positions to joints using the PositionJointInterface. It's stable and deterministic, ideal for precision motion in known environments.

• Demo Behavior: Joints followed a smooth cosine-profiled offset starting from a known initial configuration.

• Pros: Highly precise; easy to implement.

• Cons: Zero compliance; reacts poorly to disturbances.

• Applications: Calibration, trajectory playback, repeatable motion tasks. 

3. Joint Impedance Controller

Implements a spring-damper behavior at each joint using EffortJointInterface for torque control. It responds to position/velocity errors, offering compliance and safety.

• Demo Behavior: Moved the end-effector in a circular Y-Z path using joint torques derived from desired joint trajectories.

• Pros: Compliant; ideal for interacting with environment or human.

• Cons: Requires careful gain tuning; slower response in precise tasks.

• Applications: pHRI, insertion tasks, safe manipulation, joint-level compliance. 

4. Cartesian Velocity Controller

Commands end-effector velocity in 6D Cartesian space (x, y, z, roll, pitch, yaw) using FrankaVelocityCartesianInterface.

• Demo Behavior: Robot moved diagonally in the X-Z plane at ±0.05 m/s with direction switching every 4 seconds.

• Pros: Intuitive control; ideal for teleoperation or cyclic tasks.

• Cons: No position control; drift over time.

• Applications: Surface scanning, joystick control, exploratory motion. 

5. Cartesian Pose Controller

This controller commands a full 6D target pose for the end-effector using the FrankaPoseCartesianInterface.

• Demo Behavior: Used a smooth, time-varying double-cosine profile to generate displacement trajectories in Cartesian space. The controller follows the waypoints received from Haply Inverse3.

• Pros: Clean abstraction of goal-driven motion in workspace; ideal for task-level planning.

• Cons: Not contact-aware; can behave poorly near singularities or under external forces.

• Applications: Waypoint tracking, teach-by-demonstration, pick-and-place in structured space. 

6. Cartesian Impedance Controller

Creates a virtual mass-spring-damper system in Cartesian space. Computes forces based on pose error and maps them to joint torques using Jacobians.

• Demo Behavior: Trajectory followed while staying compliant to external forces. Torque commands limited for safety. The controller follows the waypoints received from Haply Inverse3.

• Pros: Balances precision and compliance; safe in HRI; highly tunable.

• Cons: Sensitive to parameter tuning and orientation errors.

• Applications: Assembly, human collaboration, safe exploration, adaptive surface interaction. 

7. Force Controller

Maintains a desired contact force using torque sensors and closed-loop feedback. Forces are translated to torques using Jacobian transpose.

• Demo Behavior: End-effector pressed downward with a constant virtual mass-derived force. PI control adjusted torques in real-time. Here mass = 0.

• Pros: Precise control over interaction forces; simple physics-based model.

• Cons: Sensitive to modeling errors, sensor drift, and contact loss.

• Applications: Polishing, force-based exploration, compliant pressing or insertion. 

Key Results

• Best Overall: Cartesian Impedance Control — most suitable for dynamic and contact-rich tasks.

• Most Precise: Joint Position Controller — excellent for calibration and structured environments.

• Best for Force Tasks: Force Controller — when accurate contact force is critical.

• Best for Smooth Motion: Cartesian Velocity Controller — for teleoperation or path scanning.