Model predictive control (MPC) has become popular in autonomous control.

The key is to optimize control actions over a prediction horizon, which is often short due to computation.

It is however desirable for AVs to have long prediction, hundreds meters or more like sensor/map ranges, so that it can plan well in advance when avoiding collision, approaching intersections/roundabout,… and optimize vehicle performance, energy consumption altogether.

We developed an MPC to ensure safety within an infinite horizon formally using control barrier function (CBF). With long road prediction in mind, the car can enhance comfort driving as well, i.e. smooth rather than sudden acceleration/ braking.

Please check here (https://lnkd.in/d5QpUzh) the paper for detail. And a short comparison video :) .