Model-based control techniques are widely used for autonomous navigation. MPC, in particular, is a flexible and powerful technique that allows considering different navigation objectives and constraints and performing trajectory tracking and obstacle avoidance at the same time. Furthermore, the MPC can be equipped with a model of the robot/vehicle that is continuously updated/adapted to changing conditions of the environment, using learning techniques, also taking into account the level of uncertainty associated with the model itself.

This thesis aims to develop an MPC controller for the autonomous navigation of a tracked robot. The controller will be based on a simple model of the vehicle/robot and on adequate algorithms to adapt it to changes in the environment. Furthermore, the controller will be aimed at autonomous navigation in agricultural environments, such as navigation in rows and open fields.
The developed controller will be validated in simulation and/or in the field using an Agilex Bunker or Bunker Pro tracked robot.

This work is suitable for both students coming from an engineering and non-engineering background, and can be tackled both by a single student and by a group of two students.

Reference teacher: Prof. Luca Bascetta – luca.bascetta@polimi.it