The goal for the ConACon project is to develop tools to assist companies to make their controllers adaptive. We aim to do this for slowly varying contexts that can be robustly estimated, and based on these contexts we will adapt the parameters of the currently used control laws in a precomputed manner. As such, a robust and easy-to-implement adaptive controller is obtained, with can be applied in industrial cases. In line with the industrial use cases, we will limit the scope to:

• Slowly varying contexts which change on average at least 10x slower than the normal control loop, consisting of known, measured or estimated operating conditions, system states, disturbances, executed tasks, setpoints, operator inputs, and indirectly derived quantities.
• Cases with a relatively low number of contexts (up to 20).
• Cases where the needed number of sensors to determine the context is relatively low (up to 20), although we may have to choose these from a bigger list of available sensors. Adapting parameters of currently used control laws; these will be parameters for feedforward controllers affecting shape and/or timing, supervisory controllers selecting setpoints for lower level feedback controllers, and sometimes also the low-level feedback controllers whose parameters are adjusted.

To enable this we will need to develop:

• a toolchain and range of methods to develop a context aware controller offline, before deployment. This toolchain yields as an output a set of control parameter values as function of context, which when combined with a context estimator can be implemented and run online on current control hardware;
• an additional toolchain to efficiently validate and test the resulting controller;

Both toolchains should be based on common engineering languages such as MATLAB and/or Python. These partners are present in the project: Dana, Punch Powertrain, Ivex, TML, SISW, Picanol.