Hierarchical PID Controller Design

for a Flexible Link Robot Manipulator

Using The Bees Algorithm

D. T. Pham, Ebubekir Koç, Mete Kalyoncu, Mustafa Tınkır

Presented by Ebubekir Koç

Manufacturing Engineering Centre (MEC)
Cardiff University, UK
E-mail: [email protected]
Outline

• Introduction
• The Bees Algorithm
1. The foraging process in nature
2. Proposed Bees Algorithm
3. A Simple Example

• Experiments
• PID Controller Design for a Flexible Link
Robot Manipulator
• Applications
 Introduction

• A novel tool for complex optimisation problems.

• Inspired by the food foraging behaviour of honey bees.

• Neighbourhood search + Random search

• Selectively explores the most promising flower patches

looking for the point of maximum fitness function.
The Bees Algorithm 1. The foraging process in nature

• Scout bees search for food randomly from one flower site to another.
• The bees evaluate the different sites according to:
- The quality of the food
- The amount of energy usage

Hive

A
C

B
The Bees Algorithm 1. The foraging process in nature

• The dancer bees meet other bees at the dance floor with high
population density.

Dance floor Beehive

 The Bees Algorithm 1. The foraging process in nature

• Bees communicate through a waggle dance which contains information

about:
1. The direction of flower patches (Angle between the sun and
patch)
2. The distance from the hive (Duration of the dance)
3. The quality rating (Frequency of the dance)

Waggle Dance Movie

The Bees Algorithm 1. The foraging process in nature

• According to the fitness, sites can be visited by

more bees or abandoned.

Hive
A
C

B
The Bees Algorithm 2. Proposed Bees Algorithm

Initialise a Population of n Scout Bees

Evaluate the Fitness of the Population

Neighbourhood Search
Select m Sites for Neighbourhood
Search

Recruit Bees for Selected Sites

Proposed Bees Algorithm (More bees for the elite Sites)

Select the Fittest Bees from Each Site

Abandon Sites without New Information

Assign the (n–m) Remaining Bees to

Random Search

New Population of Scout Bees

The Bees Algorithm 3. A Simple Example

Maximum
y

*
* *
* *
*
* * * *
x

Figure 1. Initialise a population of (n=10) scout bees

with random search and evaluate the fitness.
The Bees Algorithm 3. A Simple Example

y
Hive

▪
▪ ▪
▪ ▪
*
* * * *
x

Figure 2. Select best (m=5) sites for neighbourhood search.

The Bees Algorithm 3. A Simple Example

▪
▪ ▪
▪ ▪

Figure 3. Determine the size of neighbourhood.

The Bees Algorithm 3. A Simple Example

y
Hive

**
*
▪
*
*
▪
*
*
*
▪
*
* *
▪ ▪
*

* *

Figure 4. Recruit bees for selected sites

(Number of bees around each selected site: e = 4 and m-e = 2).
The Bees Algorithm 3. A Simple Example

**
* *▪
*
▪
*
*
*
▪*
*
*
▪ ▪
*

* *

Figure 5. Select the fittest bee “ * ” from each site.

The Bees Algorithm 3. A Simple Example

y Hive

** * * *
*
* *
* *

Figure 6. Assign the (n–m) remaining bees to random search.

The Bees Algorithm 3. A Simple Example

y
*

* *
*
*

Figure 7. Find the global best point.

Experiments
Test Functions Test Functions

Table 1: Test function

Experiments
Results Test Functions & Results

Table 2: Result
-SIMPSA - Simplex method
-NE SIMPSA -The stochastic simulated annealing optimisation procedure
-GA-the Genetic algorithm
-ANTS-Ant colony system
Experiments Test Functions & Results

Custom Made Benchmark Functions

* PSO Evolutionary Alg. Artificial BC Bees Algorithm

Test function 1 x x

Test function 2 x

Test function 3 x

Test function 4 x

Test function 5 x

Test function 6 x

Test function 7 x

Test function 8 x

Test function 9 x

Test function 10 x

*Table and functions by Marco Castellani

Robot & Dynamic Modelling

Schematic
Diagram of the
Robot

Flexible arm


T
X

Hub
Robot & Dynamic Modelling

•Dynamic equation is represented in matrix form as:

Mq  Cq  Kq  F

•The state-space form of the system can be obtained as:

  0 0 1 0    0 
  
  0 0 0 1    0 

      Vm
  0 566.46  37.02 0    65.11 
  0  921.77 37.02   
  0    65.11

•The objective function is:

J  w 1.t d  w 2 .t r  w 3 .t p  w 4 .M  w 5 .ess  w 6 . A max
Results (Angular Displacement Control)
Results (Tip Deflection Control)
Results (Servomotor Control Voltages of the Controllers )
Applications
• Functions optimisation.
• Training neural networks for pattern recognition (MLP, LVQ, RBF,
SVM).
• Cellular Manufacturing.
• Job Scheduling.
• Finding multiple feasible solutions to a preliminary design
problems.
• Data clustering.
• Optimising the design of mechanical components.
• Multi-Objective Optimisation.
• Tuning a fuzzy logic controller for a robot gymnast.
• Fuzzy Clustering.
• PCB Assembly Optimisation.
• Feature Selection for Manufacturing Data.
• Generating Branded Product Concept.
Questions/Comments

Comments and Questions

[email protected]