Non Linear Control of Four Wheel

Omnidirectional Mobile Robot:

Modeling, Simulation and Real-Time
Implementation
Veer Alakshendra Dr Shital S.Chiddarwar
Research Scholar Supervisor
Robotics Lab Robotics Lab

Department of Mechanical Engineering

Visvesvaraya National Institute of Technology, Nagpur
India-440010
Contents
• Introduction
• Objective
• How Matlab/Simulink helped us?
• Mathematical Modelling
• Controller Design
• Optimization
• Simulation Results
• Real Time Implementation (RTI)
• Conclusions
• Summary

Visvesvaraya National Institute of Technology, Nagpur 2

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Wheeled mobile robots

Non holonomic : Controllable D.O.F < Total D.OF
2

Direction of
1 motion is restricted

Holonomic : Controllable D.O.F = Total D.OF

Direction of motion is not

restricted. The robot can
move in any direction.

Visvesvaraya National Institute of Technology, Nagpur 3

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Four wheel omnidirectional mobile robot (FWOMR)

• Omni-directional wheel platform is a holonomic system
having 3 DOF in horizontal plane.
Mecanum Wheel

Direction of
motion

Motor

Visvesvaraya National Institute of Technology, Nagpur 4

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Applications of FWOMR

Omni-directional wheel chair Omni fork lifter Warehouse

Omni move lifter Omni robot Tokomak building

Visvesvaraya National Institute of Technology, Nagpur 5

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Objective

Development of mathematical model of

Mecanum wheeled mobile robot.
Solution of highly non linear equations in
less computational time.
Development of a simulation environment.
Development of robust controller for
trajectory tracking of mobile robot in
presence of matched and unmatched
uncertainties.
Real time implementation of proposed
approach for trajectory tracking.

Visvesvaraya National Institute of Technology, Nagpur 6

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Work Flow

Modeling

Controller
Design

Optimization

Real-Time
Testing

Visvesvaraya National Institute of Technology, Nagpur 7

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Challenges
faced Solution
Solving second order non Simulink
linear differential equation.
Estimating control design Simulink Design
parameters. Optimization
Building a simulation
environment. Simscape
Multibody
Feedback data collection.
Real time Kinect Support
implementation on package
hardware.
Arduino
Support
Package

Robotics System
Toolbox

Visvesvaraya National Institute of Technology, Nagpur 8

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Modeling Tool Used

Simulink
Simscape Multibody

Kinect Support
package
Simulink Design
Optimization

Robotics System Arduino Support

Toolbox Package

Visvesvaraya National Institute of Technology, Nagpur 9

 Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Kinematic Equation
Vy 2  VMW 2  Vr 2 / 2  v y  a .wz Vy1  VMW 1  Vr1 / 2  v y  a .wz

Vx 2   Vr 2 / 2  vx  b .wz Vx1  Vr1 / 2  vx  b .wz

Vy 3  VMW 3  Vr 3 / 2  v y  a .wz Vy 4  VMW 4  Vr 4 / 2  v y  a .wz

Vx3   Vr 3 / 2  vx  b .wz Vx 4   Vr 4 / 2  vx  b .wz

1  yr
 x
v   1 1  1 1   cos( )  sin( ) 0 
  R
   2  R( )   sin( ) cos( ) 0 
 y
v

1 1 1 1
 
xr
w  4  1 /(a  b) 1 /(a  b) 1 /(a  b)  1 /(a  b) 3   0 1 

 z    0
3 
yq

   R( )  xr z 
T
Pq   xq
T
yq yr
xq

Visvesvaraya National Institute of Technology, Nagpur 10

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Dynamic equation

x(t )  f ( x)  g ( x)u(t )   (t , u(t ))

Symbolic math toolbox

Visvesvaraya National Institute of Technology, Nagpur 11

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Solution of dynamic equation

Simscape Multibody
Equation of motion can be solved by either solving the derived
equation using Simulink or by using Simscape Multibody.
In Simscape Multibody , CAD file of the robot is imported to
Matlab in the form of masses, inertias, sensors, joints etc. It also
provides a 3D environment to visualize the system dynamics.
To check the system dynamics, input voltage can be kept variable
and it can be manually changed using Slider block available in
the Dashboard library.
Alternatively, Signal builder block can be used to build the input
signal.

Visvesvaraya National Institute of Technology, Nagpur 12

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Second order sliding mode control

error  as t  0 s (t )  k p e(t )  kd e(t )  ki e(t )   s(t )

s (t )  0
utotal  uequivalent  uswitching

uequivalent  (kd B34 ) 1 (ki e(t )31  k p e(t )31

 kd ( xd (t )31  Ax(t )31
 Bx(t )31 )   s(t )31 )
uswitching   s(t )31  k sw sat ( s (t )31 )

Lyapunov
 kd 1 0 0  k p1 0 0  ki1 0 0
stability
kd   0 kd 2 0  k p   0 kp2 0
 ki   0 ki 2 0 
criterion is
 0 0 kd 3  0 0 k p 3   0 0 ki 3 
 satisfied
 1 0 0   ksw1 0 0  1 0 0 V 0
   0  2 0  ksw   0 ksw 2 0     0 2 0 

 0 0  3   0 0 ksw3   0 0 3 

Visvesvaraya National Institute of Technology, Nagpur 13

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Second order sliding mode control in simulink

Dynamics

Controller
Input/output and
uncertainties

Visvesvaraya National Institute of Technology, Nagpur 14

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Simulink Design Optimization has

been used to estimate the controller
design parameters

Visvesvaraya National Institute of Technology, Nagpur 15

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Simulation results for complicated trajectory

Visvesvaraya National Institute of Technology, Nagpur 16

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Open loop testing

Visvesvaraya National Institute of Technology, Nagpur 17

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Close loop testing (Robotics System Toolbox)

Node Node

Kinect ROS
Matlab/Simulink Node
program

Simulink
ROS
IMU
Matlab/Simulink
program

OR

Kinect ROS
package

IMU ROS Node

package
ROS

Visvesvaraya National Institute of Technology, Nagpur 18

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Close loop testing (Robotics System Toolbox)

Subscriber Controller Publisher

Visvesvaraya National Institute of Technology, Nagpur 19

Introduction Objective How Matlab helped us Mathematical Modelling Controller Design Optimization Simulation Results RTI

Results for eight type trajectory

Visvesvaraya National Institute of Technology, Nagpur 20

Conclusions
• Solution of non linear dynamic equation including
orientation of mobile robot.
• Successful tracking of desired trajectory in presence of
matched and unmatched uncertainties.
• Smooth and chattering free control input.
• Successful real time implementation.

Visvesvaraya National Institute of Technology, Nagpur 21

 Summary of the work and advantages of using
Matlab
Summary Advantages of using Matlab

Closed loop test using

camera and IMU. Better
Model based design understanding of
kinematic and
Open loop test using dynamic model
Arduino MEGA ADK Implementation of
control strategy
Implementation of
Estimation of control control strategy
design parameters Ability to solve non
linear equations
Ability to perform
number of
Control Law simulation
Parameter
experiments
estimation
Dynamic Equation
Real time
implementation
Kinematic Equation

Visvesvaraya National Institute of Technology, Nagpur 22

THANK YOU

Q & A?

Visvesvaraya National Institute of Technology, Nagpur 23