Control kernel based adaptive control implementation

Abstract

A control system with distributed computing resources always should guarantee the safe control of the plant. In this contribution, the concept of control kernel is used for that purpose. Two types of nodes with different resources are defined: the powerful server node and the resource-constrained light node. This architecture allows to split the control tasks into two blocks. Those demanding strong computing resources are allocated in the server nodes and those compelling tasks required to ensure the safety of the controlled plant are allocated in the light nodes. Resource limitations lead to control adaptation. Two simple applications illustrate some of the benefits of this architecture with one server node and one light node, even the architecture can be extended to several nodes. In the first case, an adaptive control is implemented in the server node, providing the control algorithm to the light node, which is also able to compute a local safe control action. In the second experiment, two different control tasks requiring different resources are implemented in a mobile robot control. To keep bounded the computing time at the local level, the supervisor decides the time allocated to each activity, providing the resulting controller to the light node.