A software tool for code generation of linear model predictive control for real-time embedded applications.

The tool website is found here.

A 2-link robotic arm with constraints in the inputs and states.

Fast Online MPC: The optimal control problem is in the form of a quadratic program, which is solved using augmented Lagrange multipliers together with Nesterov's gradient method.
System: a linear plant with 4 states, 2 inputs, horizon lengths of 5 steps.
Hardware: 48 MHz ARM7 microcontroller. Around 16 kB of memory have been used for the compiled code.
Software: nxtOSEK real-time operating system, with 4 ms hard real-time MPC task.

Online MPC for fast embedded mechatronic systems -
Top part of the video: constraints on states and inputs, bottom: constraints on the inputs only.

Go to Top A Segway-like two-wheel vehicle. We use a linear MPC scheme to stabilise the system.

Fast Online MPC: The optimal control problem is in the form of a quadratic program, which is solved using Nesterov's gradient method.
System: a linear plant with 8 states, 2 inputs, horizon lengths up to 20 steps.
Hardware: 48 MHz ARM7 microcontroller. Around 8 kB of memory have been used for the variables of the algorithm.
Software: nxtOSEK real-time operating system, with 4 ms hard real-time MPC task.
Publication Preprint of article on ACC 2012: available here.

Online MPC for fast embedded mechatronic systems -
Input constrained case.

Go to Top