Miriam Hoekstra*, Mathijs Vogelzang, Evgeny Verbitskiy and Maarten WN Nijsten
Corresponding author: Miriam Hoekstra email@example.com
Critical Care 2009, 13:223 doi:10.1186/cc8023
B. Wayne Bequette
(2012-01-03 11:19) Rensselaer Polytechnic Institute
Hoekstra and coworkers  review the technologies available for computerized glucose
regulation in the intensive care unit, but misrepresent the differences between two
control algorithms, Proportional-Integral-Derivative (PID) and Model Predictive Control
(MPC). The differences between PID and MPC are illustrated by an example of an automobile
on a roadway. They claim that the driver using MPC determines his/her driving strategy
before departing, and maintains that trajectory throughout the trip. They also claim
that the driver using PID makes frequent control action changes based on the difference
between the ��ideal�� and actual trajectory. The MPC scenario shown is largely incorrect.
MPC looks into the future (down the roadway) and determines the best sequence of control
actions (driving strategy) to maintain that future trajectory. MPC does not simply
implement that entire sequence of control actions (steering, braking, etc.), but,
instead, updates the control actions at frequent intervals, in the same way that PID
makes adjustments at frequent time intervals. Thus, the MPC strategy is no more sensitive
to ��small errors in input variables�� than the PID strategy.
It has been a common misconception that PID is less sensitive than MPC to uncertainty
in the system dynamics. In reality, MPC is no more sensitive to uncertainty than PID,
if the two strategies are tuned for the same performance. Undergraduate textbooks
 show that internal model control (IMC), a model-based control strategy, can be
implemented as an equivalent PID controller when low-order models are used as the
basis for controller design; this is also shown by Percival and co-workers . The
��take home message�� is, if PID and MPC algorithms are tuned for the same level of
performance, they have exactly the same sensitivity to uncertainty.
1. Hoekstra M, Vogelzang M, Verbitsky E, Nijsten MWN: Health technology assessment
review: Computerized glucose regulation in the intensive care unit �� how to create
artificial control. Critical Care 2009, 13:223 (doi:10.1186/cc8023)
2. Bequette BW: Process Control: Modeling, Design and Simulation. New Jersey: Prentice
3. Percival MW, Zisser H, Jovanovic L, Doyle III FJ: Closed-loop control and advisory
mode evaluation of an artificial pancreatic b cell: Use of proportional-integral-derivative
equivalent model-based controllers. J. Diabetes Sci Technol 2008, 2:636-644.
BioMed Central Ltd unless otherwise stated. Part of Springer Science+Business Media.