M A HAFIZ

Sim2Real & Control Architecture

How I structure robotics systems for stable control + measurable evaluation

System Overview

My workflow is designed for repeatability: human input (gesture/keyboard) → control logic → simulator/hardware → logging → evaluation → regression gating. This keeps iteration fast and measurable.

Results Projects Contact
High-level pipeline (edit this to match your exact stack): 
[EMG/IMU] -> [Gesture Classifier] -> [Menu/State Machine]
                                -> [Command (Axis + +/-)]
                                -> [Controller (Cartesian/IK/Impedance)]
                                -> [MuJoCo Simulation] <-> [State Feedback]
                                -> [Logger] -> [Metrics + Video + Report]
                                -> [Regression Gate: SHIP/BLOCK]

Core Blocks

1) Input + Intent

  • Gesture classification from EMG/IMU (or keyboard for debugging)
  • Menu-driven UX to select axis/mode (X/Y/Z, Gripper, Screw/Unscrew)
  • Safety: lock/unlock, rate limiting, dead-man behavior
HRIState MachineSafety

2) Control Layer

  • Cartesian delta commands converted into joint updates (Jacobian / IK)
  • Optional impedance behavior for compliant interaction
  • Orientation handling: keep fixed or allow controlled rotation
Jacobian IKCartesian ControlImpedance

3) Plant (Sim / Real)

  • MuJoCo scene matches lab setup (table, objects, contacts)
  • Franka Panda (FR3) execution for sim2real validation
  • Consistent interfaces: same command format & feedback structure
MuJoCoFranka FR3Sim2Real

4) Evaluation & Regression Gating

  • Log episodes (states, commands, success/failure reasons)
  • Generate metrics + videos automatically
  • Compare baseline vs candidate: SHIP/BLOCK decision
MetricsCI/CDReproducibility

Stability & Safety Strategies I Use

  • Command bounding: clamp Δx/Δy/Δz and Δrot per step
  • Rate limiting: cap command frequency and smooth inputs
  • Hold behavior: keep pose / joint targets to prevent sagging
  • Failure visibility: log “why it failed” (stale, instability, collision, drift)
  • Recovery: safe reset + controlled re-approach to target

Add diagrams & media (recommended)

Add 2–3 visuals here. Even simple screenshots improve credibility:

  • Block diagram (pipeline)
  • MuJoCo scene screenshot + real robot photo
  • Plot: desired vs actual pose
Placeholder: Put images in images/ and embed like: 
<img src="images/architecture_diagram.png" alt="Architecture Diagram" style="width:100%; border-radius:12px; border:1px solid #ddd;">