Immune Genetic Algorithms for Optimization of Task Priorities and FlexRay Frame Identifiers

1
Immune Genetic Algorithms for Optimization of
Task Priorities and FlexRay Frame Identifiers

• Soheil Samii1, Yanfei Yin1,2, Zebo Peng1, Petru
  Eles1, Yuanping Zhang2

1 Dept. of Computer and Information
Science Linköping University Sweden
2 School of Computer and Communication Lanzhou
University China
2
Motivation

• FlexRay
• Safety-critical applications in the static
  segment
• Other applications in the dynamic segment

Many optimization parameters
3
Outline

• System model
• Bus cycle of FlexRay
• Problem formulation
• Optimization with immune genetic algorithms
• Experimental results

4
System model
5
FlexRay configuration
Static phase
Dynamic phase
a
c
d
b
2
3
1
2
3
Bus cycle
1

• Frame identifiers and priorities to messages
• Priorities to tasks

a
c
d
b
1
3
2
6
Timing with some configuration
Average 477
7
Timing with some other configuration
Average 369
Previous case Average 477
8
Problem formulation

• Parameters
• Priorities of the tasks
• Frame identifiers and priorities of the messages
• Objective
• Minimize the average response time of tasks

9
Immune genetic algorithms
Crossover
Initial population
Simulation of each member
Mutation
Evaluate costs
Population
Population costs
Vaccination
Stop?
New population
No
10
Vaccination
Population
Create vaccines
4
2
3
4
2
1
Dominance threshold 50
60
80
80
40
60
60
11
Vaccination
Population
Vaccination rate
Create vaccines
Select member
Vaccine set
Member
Select vaccines
Vaccines
Vaccinate
12
Vaccination
Member
Vaccines
New member
13
Vaccination
Population
Create vaccines
Vaccination rate
Dominance threshold
Select member
Vaccine set
No
Member
Select vaccines
Last member?
Vaccines
Yes
Vaccinate
New population
14
Tuning Vaccination rate
Cost
Vaccination rate
15
Tuning Dominance threshold
Cost
Dominance threshold
16
Vaccination

• Takes advantage of local properties of good
  solutions
• Speed up the optimization process
• Improve the quality of the final solution

17
Experiments Improvements
Cost improvements
Number of tasks
18
Experiments Runtime
Runtimes seconds
Number of tasks
19
Conclusions

• Minimize delays in distributed embedded systems
• Task priorities
• Frame identifiers for FlexRay messages
• Immune genetic algorithms
• Vaccination results in better optimization in
  terms of time and solution quality (compared to
  traditional genetic algorithms)