2-DOF Manipulator — Model-Based Real-Time Control

A MATLAB/Simulink + TI C2000 control stack that combines computed-torque, PD + gravity compensation, and joint/task-space transformations to track millimetre-level trajectories on a 5-link planar robot in real time.

1. Overview & Motivation

This project—completed for ME403 : Introduction to Robotics (Spring 2024, Sabancı University)—developed a model-based real-time controller for a planar 2-DOF, 5-link manipulator.
Key goals:

  • Achieve sub-millimetre end-effector accuracy on dynamic paths
  • Bridge simulation (SIL) and embedded deployment (HIL) seamlessly
  • Provide a re-usable MATLAB/Simulink template for undergraduate robotics labs


CAD and joint assignment

Figure – CAD view and joint layout of the 5-link manipulator.


2. Control Architecture

The stack follows a computed-torque framework augmented with a PD + gravity compensator.
Software is split into three Simulink model hierarchies that map 1-to-1 onto TI C2000 peripherals.

Layer Role Runs On
Task-Space Planner Generates x-y trajectories Host PC (SIL/HIL)
Controller Core Inverse kinematics + computed-torque PD TI C2000 (HIL)
Motor Interface PWM output + encoder decoding TI C2000 (HIL)

Control Loop (1 kHz)

  1. Trajectory Sample – desired end-effector pose
  2. Inverse Kinematics – joint targets (θdes, ẋdes)
  3. Computed-Torque Law – τ = M(q)[*]α + V + G
  4. PD + G Compensation – fine-tunes stiffness & damping
  5. PWM Generation – duty cycles to H-bridges
  6. Encoder Feedback – closes joint-space loop
% Simulink MATLAB Function: Computed-Torque PD
tau = M(q)*(qdd_des + Kd*(qd_des - qd) + Kp*(q_des - q)) ...
      + V(q, qd) + G(q);


3. Simulation & Real-Time Testing

Stage Environment Purpose
SIL Simscape Multibody + MATLAB ODE45 Verify kinematics, dynamics, and controller logic without I/O latency
Processor-in-the-Loop TI C2000 F28379D + external mode Measure computation time (< 180 µs / step)
HIL Full manipulator with DC motors & optical encoders Validate closed-loop tracking and load tolerance

Total transition time from SIL to validated HIL: < 3 hours thanks to code-gen-friendly model structure.


4. Results & Performance

Trajectory RMS Error Peak Error Cycle Time
Horizontal (0.4 m) 0.37 mm 1.1 mm 5 s
Vertical (0.3 m) 0.41 mm 1.3 mm 4 s
Ellipse (a = 0.25 m, b = 0.15 m) 0.62 mm 1.9 mm 8 s
  • All tests at 1 kHz loop, 24 V supply, 40 % rated torque
  • CPU load: 43 % on C2000 (180 MHz) — ample headroom for advanced features


Observations

  • Gravity term cancellation cut average torque by ≈ 22 %, reducing motor heating.
  • Controller remained stable up to 1.8 kHz; limited to 1 kHz for encoder filtering.
  • Minor elasticity in the 5-link chain explained residual ellipse error.


5. Deployment & Usage

# 1 – Generate code & flash
slbuild('manipulator_controller', 'StandaloneCoderTarget');

# 2 – Run SIL bench
sim('manipulator_system_sil.slx');

# 3 – Switch to External Mode (HIL)
manipulator_hil_start('COM5', 1e-3);  % port, sample time
  • PWM: 20 kHz, center-aligned
  • Encoders: 1024 PPR, quadrature decoded to 4096 counts/rev
  • Comms: Host-target via FTDI USB-UART @ 115200 baud


*Project by Yunus Emre Danabaş as part of ME403 — Introduction to Robotics (Spring 2024, Sabancı University).