Dynamic Environments Mathias Pacher and Uwe Brinkschulte ISORC - - PowerPoint PPT Presentation

▶
dynamic environments
SMART_READER_LITE
LIVE PREVIEW

Dynamic Environments Mathias Pacher and Uwe Brinkschulte ISORC - - PowerPoint PPT Presentation

Aug 22, 2022 164 likes •338 views

Towards an Artificial DNA for the Use in Dynamic Environments Mathias Pacher and Uwe Brinkschulte ISORC 2019, Valencia May 9, 2019 1. Motivation Current ICT systems : Increasingly complex Distributed Interconnected Dynamic

slide-1
SLIDE 1

Towards an Artificial DNA for the Use in Dynamic Environments

Mathias Pacher and Uwe Brinkschulte

ISORC 2019, Valencia May 9, 2019

slide-2
SLIDE 2
  • 1. Motivation

Current ICT systems:

  • Increasingly complex
  • Distributed
  • Interconnected
  • Dynamic environments

➔Thus,

  • Development and maintenace are hard
  • Failures at run-time

Idee of Organic Computing:

  • System adapts autonomously and dynamically to environment

(Tomforde et al., „Organic Computing in the Spotlight“, 2017)

1

slide-3
SLIDE 3
  • 2. Artificial Hormone System
  • Assignment of tasks to processors
  • Hormone-based control loops
  • Self-configuration
  • Self-improvement
  • Self-healing
  • Create virtual organs

2

Middleware Tasks

PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ PZ

Organs

Application

Brinkschulte, Pacher, von Renteln, An Artificial Hormone System for Self-Organizing Real-Time Task Allocation in Organic Middleware, Springer

slide-4
SLIDE 4
  • 3. Artificial DNA

3

Idea:

  • Most embedded systems consist of standard components

➔ Describe components and interconnection as a text file ➔ Artificial DNA ➔ No programming, only parametrization ➔ Automatically determine tasks and hormone strength

1 = 70 (1:2.2) 100 25 // constant setpoint value, period 25 msec 2 = 1 (1:3.1) ‐ // ALU, control deviation (minus) 3 = 10 (1:4.1) 4 5 6 25 // PID (4, 5, 6), period 25 msec 4 = 600 1 // actor, resource id = 1 5 = 500 (1:2.1) 2 25 // sensor, resource id = 2, period 25 msec

slide-5
SLIDE 5
  • 4. Dependability

4

➔ Interesting approach for automotive applications In general: 𝑄

𝐵𝐸𝑂𝐵 ≤ 𝑄𝑆𝑓𝑒.

slide-6
SLIDE 6
  • 5. Artificial DNA for dynamic environments
  • Self-building system at run-time
  • Easy to configure at run-time
  • Scalable
  • ADNAs of different systems may merge and separate

5

slide-7
SLIDE 7

Example

6

Red car:

  • ABS 1
  • ESP 1
  • Motor control 1
  • Entertainment 1
  • …

Blue car (less computing power):

  • ABS 2
  • ESP 2
  • Motor control 2
  • Entertainment 2
  • …
slide-8
SLIDE 8

Example

7

Scenarios: 1. Stress test → Different car is in range each 1.5 seconds for 1.5 seconds 2. Replacement for failing processors → Different car is in range each 6 seconds for 4.5 seconds

slide-9
SLIDE 9

Evaluation scenario 1 (stress test)

8

  • ADNA of red car complete
slide-10
SLIDE 10

Evaluation scenario 1 (stress test)

9

  • Reaction time until ADNA is complete
slide-11
SLIDE 11

Evaluation scenario 1 (stress test)

10

  • Speed and distance
slide-12
SLIDE 12

Evaluation scenario 2 (Replacement for failing processors)

11

  • ADNA of red car complete
slide-13
SLIDE 13

Evaluation scenario 2 (Replacement for failing processors)

12

  • Reaction time until ADNA is complete
slide-14
SLIDE 14

Evaluation scenario 2 (Replacement for failing processors)

13

  • Speed and distance
slide-15
SLIDE 15

Conclusion

  • First experiments with extended ADNA for dynamic systems
  • Stress test
  • Compensate failing processors

Future work:

  • ADNA assignment priorities
  • Conditional ADNA

Then:

  • Paywall for automotive applications

14

Conditional part Unconditional part

slide-16
SLIDE 16

Thank you

Questions?

15