Dynamic balancing in the real world with GRiSP

Details

Supervisors

Van Roy, Peter

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil électromécanicien, à finalité spécialisée: mécatronique

Abstract

Unstable systems are present in many engineering domains, such as industrial plants, energy production, aeronautics, transportation and medical. Such systems are really challenging to control in order to achieve specific missions, such as controlling complex chemical processes, putting a satellite into orbit, restarting a human heart, or even deploying legged or self-balancing robots. It has been, and still is, at the heart of many innovations. The objective of this master thesis is to control an unstable system using a GRiSP board. This compact circuit board features a wide range of ports and connectivity while running under Erlang, a multi-paradigm language that offers a number of features of great interest in the world of control. This master thesis builds on top of the Hera framework developed over several past master theses to enable the use of Kalman filter-based sensor fusion, a very powerful tool for measurement enhancement. As a concrete, yet representative and generalisable use case, a real-world unstable system was selected and built: the two-wheeled self-balancing robot. After designing such a device to be compatible with GRiSP, a control strategy was established to make it dynamically balanced. The software was then created and optimised to accommodate this theoretical control loop and the various associated sub-components. A Kalman filter, fed by multiple sensors, was implemented, along with a simpler complementary filter, allowing for performance comparison. Additional features that make this robot stand out are its lifting mechanism and the possibility of being remotely controlled. This system is multilevel with two levels: an outer executive loop deciding how the robot moves around, and an inner stability loop ensuring stability. This results in the creation of a totally autonomous robot capable of positioning itself in its balancing state. This system is very powerful and reliable. It is able to move agilely while withstanding most external disturbances that were experimentally characterised and compared with other solutions. The whole architecture supports the development of similar systems in various fields, such as robotics, automation or IoT, through the reuse and adaptation of parts like the outer state machine loop or the inner control loop. It is also well documented and comes with a set of development and debugging tools.