Title: Evolvable Fuzzy Hardware for Real-time Embedded Control in Packet Switching
1Evolvable Fuzzy Hardware for Real-time Embedded
Control in Packet Switching
- Ju Hui Li
- Meng Hiot Lim
- Qi Cao
2Outline
Nanyang Technological University, Singapore
- Introduction to EHW
- Evolvable Fuzzy Hardware (EFH)
- Hardware Implementation (RFIC)
- ATM Cell-Scheduling
- EFH application on Cell-Scheduling
- Simulation Results
- Conclusion and Future Work
3Introduction to EHW
Nanyang Technological University, Singapore
- Definition
- Modify
- Autonomously
- Classification Methods
- Methods of Evolution
- Extrinsic, Intrinsic
- Adaptation Scheme
- On-line, Off-line
- Evolutionary Granularity
- Transistor, Gate, Function Units
4Introduction to EHW
Nanyang Technological University, Singapore
- Open Issues
- On-line Adaptation
- Scalability
- Termination of Evolution
5Evolvable Fuzzy Hardware
Nanyang Technological University, Singapore
- Evolvable Fuzzy Hardware (EFH)
- Traditional Fuzzy Hardware
- Fuzzy Rule Set Designed by Experts
- Considering The Whole Scenario
- Fuzzy Rule Set Fixed
- EFH
- Fuzzy Rule Set Searched by GA
- The Small Period Scenario
- Fuzzy Rule Set may change
6Evolvable Fuzzy Hardware
Nanyang Technological University, Singapore
Evaluation
7Evolvable Fuzzy Hardware
Nanyang Technological University, Singapore
- Evolution Scheme
- Training Data
- Pattern Prediction
- Search for a good but not optimal rule set
- The baseline is the working rule set
- If no better chromosome can be found within the
fixed generations, the working fuzzy rule set is
deemed to be good enough - Core rule set
8Evolvable Fuzzy Hardware
Nanyang Technological University, Singapore
- RFIC (Reconfigurable Fuzzy Inference Chip)
9ATM Cell-Scheduling Problem
Nanyang Technological University, Singapore
- Problem Description
- Class1
- Class2
- The capacity of Channels
10ATM Cell-Scheduling Problem
Nanyang Technological University, Singapore
- Quality of Service
- Class1 Cell Delay
- Cell Loss (Class1 and Class2)
- Balance of Cell Losses (Fairness)
11ATM Cell-Scheduling Problem ---Available
Schemes
Nanyang Technological University, Singapore
- FIFO
- DWPS
- Other Methods
- Round Robin Scheduling
- Generalized Processor Sharing
12EFH Application on Cell-Scheduling
Nanyang Technological University, Singapore
Training buffer1 (TB1)
Training buffer2 (TB2)
13EFH application on Cell-Scheduling
Nanyang Technological University, Singapore
- Training Data
- Principle of Locality
- Using Past Period Cell Flow to Train EFH
- Search for a good but not optimal rule set.
- The baseline is the working rule set
- If no better chromosome can be found within the
fixed generations, the working fuzzy rule set is
deemed to be good enough.
14EFH application on Cell-Scheduling
Nanyang Technological University, Singapore
- Fuzzy Variables
- C1V1/Vmax
- C2L2/Lmax
- Membership Functions
15EFH application on Cell-Scheduling
Nanyang Technological University, Singapore
- Core Rule Set
- To Prevent From Adopting Very Poor Rule Set
C1 C2 VS S M L VL
VS T F F F F
S T T T F F
M T T T F F
L T T T T F
VL T T T T T
16EFH application on Cell-Scheduling
Nanyang Technological University, Singapore
- Coding Scheme
- 12222,11122,11122,11112,11111
- Fitness Function
17EFH application on Cell-Scheduling
Nanyang Technological University, Singapore
- GA Parameters
- Generation Number9
- Evolution Cycle2
- Population Size10
- Elite Pool Size2
- Crossover Probability0.6
- Mutation Probability0.05
18EFH application on Cell-Scheduling
Nanyang Technological University, Singapore
- Simulation
- Scenario1
- CBR is 155.52MHz
- VBR is 155.52MHz
- 2 Seconds
- Scenario2
- CBR is 100MHz
- VBR is from 55.52MHz to 155.52MHz
- 2 Seconds
19Simulation Results
Nanyang Technological University, Singapore
20Nanyang Technological University, Singapore
21Nanyang Technological University, Singapore
22Simulation Results
Nanyang Technological University, Singapore
23Nanyang Technological University, Singapore
24Nanyang Technological University, Singapore
25Simulation Results
Nanyang Technological University, Singapore
- QoS Tunability of EFH
- Adjusting QoS by adjusting the value of ?.
- The smaller the ?, the smaller the class1 delay
and vice visa. - The value of ? can be decided if the desired
class1 delay is decided.
26Simulation Results
Nanyang Technological University, Singapore
27Nanyang Technological University, Singapore
28Nanyang Technological University, Singapore
29Nanyang Technological University, Singapore
30Nanyang Technological University, Singapore
31Conclusions
Nanyang Technological University, Singapore
- The proposed EFH can be successfully applied on
ATM cell scheduling - EFH can realize Intrinsic Evolution and Online
adaptation. - It can trace the flow pattern and evolve an
appropriate rule set. - It can achieve good QoS balance.
- The achieved QoS can be adjusted conveniently
32Conclusions
Nanyang Technological University, Singapore
- (E) The result is equal to or better than the
most recent human-created solution to a
long-standing problem for which there has been a
succession of increasingly better human-created
solutions. - (F) The result is equal to or better than a
result that was considered an achievement in its
field at the time it was first discovered. - (G) The result solves a problem of indisputable
difficulty in its field.