Dual Adaptive Control Toolbox (DACT)

for worst-case design of adaptive controllers

by Sándor M Veres

This toolbox is aiming at the design of high-reliability adaptive controllers acceptable by industry standards. Routines are provided to build dual-control schemes, i.e. to insure on-line interaction of identification and control. Tools are available for both stochastic and bound-based (BB) dual-control schemes with larger emphasis placed on BB-schemes because of their better potential in applications. BB-schemes can be built

(I) to provide guaranteed performance within a finite time-period.

(II) to ensure that the control action taken guarantees that on-line identification is able to track changes in plant dynamics .

The word guarantee is meant by the assumptions on I/O measurement errors and actuator errors to be met and that bounds are known on un-modelled dynamics.

The following are example m-files of simulations with worst-case dual-control schemes:

• Application example of a simple ADDITIVE CRITERION of on-line identification and control
  http://sun1.bham.ac.uk/veressm/ftparchive/d_wdadd1.m .
• Application example of a complex ADDITIVE CRITERION of on-line identification and control
  http://sun1.bham.ac.uk/veressm/ftparchive/d_wdadd2.m .
• Application example of a time-domain SYNERGETIC SCHEME of on-line identification and control
  http://sun1.bham.ac.uk/veressm/ftparchive/d_syn1.m .
• Application example of a frequency-domain SYNERGETIC SCHEME of on-line identification and control
  http://sun1.bham.ac.uk/veressm/ftparchive/d_syn2.m
• Application example of a SUBOPTIMAL DUAL CRITERION of on-line identification and control
  http://sun1.bham.ac.uk/veressm/ftparchive/d_subop1.m based on polytope-bounding.
• Application example of a SUBOPTIMAL DUAL CRITERION of on-line identification and control based on ellipsoid- bounding.

Note that to run the above demos some routines are needed from the Geometric Bounding Toolbox (GBT) available in the FTP facility

This toolbox is in preparation, further samples of routines are available on request via e-mail: s.m.veres@bham.ac.uk