acados (fast embedded optimal control solver)

An open-source solver for real-time optimal control and MPC that can generate embeddable C code.

acados is an open-source suite of software for optimal control and model predictive control (MPC — solving an optimization problem over a short future horizon at every control step), developed mainly by Moritz Diehl's group at the University of Freiburg and collaborators, and can be seen as a successor to the earlier ACADO toolkit. Its core is written in C and relies on BLASFEO (a small-matrix linear-algebra library) and HPIPM (a structured quadratic-programming solver) underneath, offering algorithms such as SQP (sequential quadratic programming) and real-time iteration (RTI), aimed at solving nonlinear MPC within a millisecond-scale control period. Users typically write system dynamics and cost functions in Python or MATLAB using CasADi, have acados generate C code from them, and then deploy that code to an industrial PC or embedded board. It's common in MPC research for legged robots, drones, and robot arms.

ExampleWrite a quadrotor's dynamics in CasADi in Python, generate a solver with acados, and run nonlinear MPC for trajectory tracking on an onboard computer at roughly 100 Hz.

Related

Model Predictive Control · Nonlinear Model Predictive Control · CasADi · Sequential Quadratic Programming · OCS2 · Crocoddyl (Contact RObot COntrol by Differential DYnamic programming Library)

Sources

acados documentation
acados/acados (GitHub)