An Object Oriented Framework For A Generic Model Predictive Controller

We present an object oriented design for a reusable nonlinear model predictive controller framework. The framework balances reusability with flexibility and performance. Separation of interface from implementation simplifies development and facilitates use of dynamic polymorphism to change the controller behaviour at runtime. The work is presented in language agnostic terms, but the success of a particular implementation for an omni-wheeled robot is briefly described. That implementation is available online, but an application non-specific version will be made available with examples as starting templates.