Model Predictive control look-ahead feature

Hi ,

I am working with a project in which i am using MPC to control my robot. I wish to know why do we provide future references to MPC to predict the future( in Simulink ) as explained in the Autonomous steering control example at mathworks website.

I am not able to understand this concept of previewing. Doesn't MPC should predict the future itself by once using one reference signal ?? But in previewing we are using set of future references to predict future in simulink.

Can someone help in explaining this concept?

Regards,

Answer 1

[Arkadiy Turevskiy]

Hello,

I'll try to explain. First, you do not need to use signal previewing. Probably in most cases you do not know how setpoint and disturbance will change in the future, so you cannon make use of "preview". However, in some cases you might know how setpoint will change in the future. For example, thermostat in a building might be programmed to increase setpoint in a predefined way after working hours. In this case, as you know in advance how setpoint will change, you may leverage that to help you get better performance.

This is explained in the doc here .

You might also want to take a look at this video that explain previewing at ~3.10.

HTH,

Arkadiy Turevskiy

MathWorks