LQ vs. ℓ_∞ in controller design for systems with delay and quantization
- Creators
- Nakahira, Yorie
Abstract
The normal operation of many cyberphysical, biological, and neural systems fit naturally with robust control, with key variables like lane positions, voltages, temperatures, blood pressures, etc maintained within tight bounds despite diverse uncertainties. However, two challenges in particularly need further theory that this paper addresses. One is that control is distributed with communication having limits on bandwidth and delay. Another is that normal operations can be disrupted and bounds violated, but it is desirable to make such acute situations rare and recoverable without crashing. We take the simplest model that has both normal and acute modes with bandwidth and delay constraints, and focus on two relatively extreme but familiar starting points: i) average case LQG (or ℋ_2) and ii) worst case ℓ_1 control with ℓ_∞ signal bounds. Both have strengths and weaknesses that we highlight, and this leads naturally to a win-win hybrid scheme that has better performance than either alone, with relatively modest computational costs.
Additional Information
© 2016 IEEE. This work was funded by grants from AFOSR and NSF, and gifts from Cisco, Huawei, and Google. We would like to thank Prof. J. C. Doyle, Prof. V. Kostina, Dr. I. Papusha, and Dr. D. Hui for insightful discussions.Attached Files
Submitted - CDC_long.pdf
Files
Name | Size | Download all |
---|---|---|
md5:9d0d79a23bb09472dea962891fc1bdd5
|
445.3 kB | Preview Download |
Additional details
- Alternative title
- LQ vs. ℓ∞ in controller design for systems with delay and quantization
- Eprint ID
- 73302
- Resolver ID
- CaltechAUTHORS:20170106-125636340
- Air Force Office of Scientific Research (AFOSR)
- NSF
- Cisco
- Huawei
- Created
-
2017-01-06Created from EPrint's datestamp field
- Updated
-
2021-11-11Created from EPrint's last_modified field