MPC is an enhanced possibilty to plan trajectories and control the vehicle. Currently we know a lot about vehicle dynamics and about model based design and therefore we can use this knowledge about the vehicle dynamics and integrate in an method called MPC. Therefore we have the possibilty to integrate nonlinear dynamic behavior e.g. from the tyres. In addition we have a fixed planning horizon for our trajectory and the control of the vehicle where with the MPC with setup a big optimization problem and solve it for our horizon. The MPC is therefore an enhanced trajectory planner for safe and high precise trajectory planning and control by including all the linear and nonlinear dynamics of the vehicle.

Besides, MPC can be seen like a real time optimization control , so with a set of advantages.

The MPC algorithm is used because a decision must be made when performing an action through probabilities in which the error is almost zero, on the other hand, the PID control simply once you have a defined value, simply stabilizes the system.

MPC offers the flexibility of specifying different objectives, not only tracking the reference (e.g., the path/trajectory) as in the case of PID. For example, in the case of autonomous driving, instead of trying to track a reference path, one may want to minimize the lateral acceleration, steering effort, etc such that the vehicle is within some bound of the path. This allows the vehicle to drive more naturally.

Model predictive control (MPC) depends on dynamical model of the plant process, it has widely been used in process control and power balancing applications.

Advantages:

Optimal control input

Desired state and input constraints can be defined

It can work well near the hard boundary and other explicit operating constraints

Ability to predict future control moves

More robust if integrated with statistical learning

Disadvantages:

Additional computational power needed and storage required to store previous commands

If prediction model is inaccurate, the system may become unstable

Difficult to obtain a reliable prediction model

Without learning can’t handle mass variations

Hi,

The MPC control recognized for its efficiency and versatility in different technological and industrial fields, it responds in a particular way to the specifications of the autonomous vehicle as well as taking charge of driving a set of vehicles with intelligent supervision having required performance.

For more information about this subject, i suggest you to see links and attached file on topic.

https://medium.com/@jonathan_hui/lane-keeping-in-autonomous-driving-with-model-predictive-control-50f06e989bc9

https://www.coursera.org/lecture/intro-self-driving-cars/lesson-4-advanced-steering-control-mpc-WHPvn

https://www.diva-portal.org/smash/get/diva2:925562/FULLTEXT01.pdf

https://pastel.archives-ouvertes.fr/tel-01635261/document

Best regards

Recently MPC is being used in real-time motion control of many systems, especially autonomous vehicles. As MPC contains a system dynamical model to predict the future states of the system, it is capable of considering environment changes in real-time. This property makes MPC capable of handling real-time motion control problems efficiently.

MPC has more realistic feature due to its dynamic modelling. We can experience the real time controls on motion.