Leader Election in a Distributed CAN-Based Multi-Microcontroller System

In a distributed system, functionally equivalent nodes work together to form a system with improved availability, reliability and fault tolerance. Thereby, the purpose is to achieve a common control objective. As multiple components cooperate to accomplish tasks, coordination between them is required. Electing a node as the temporary leader can be a possible solution to perform coordination. This work presents a self-stabilizing algorithm for the election of a leader in dynamically reconfigurable bus topology-based broadcast systems with a message and time complexity of $\mathcal{O}$(1). The election is performed dynamically, i.e., not only when the leader node fails, and is criterion-based. The criterion used is a performance related value which evaluates the properties of the node regarding the ability to perform the tasks of the leader. The increased demands on the leader are taken into account and a re-election is started when the criterion value drops below a predefined level. The goal here is to distribute the load more evenly and to reduce the probability of failure due to overload of individual nodes. For improved system availability and reduced fault rates, a management level consisting of leader, assistant and co-assistant is introduced. This reduces the number of required messages and the duration in case of non-initial election. For further reduction of required messages to uniquely determine a leader, the CAN protocol is exploited. The proposed algorithm selects a node with an improved failure rate and a reduced message and hence time complexity while satisfying the safety and termination constraints. The operation of the algorithm is validated using a hardware test setup.