Dynamic Topology Management in CAN

Abstract

The controller area network (CAN) protocol was initially developed for the automotive industry to support low-level communication services between modules in distributed car control systems. Nowadays CAN is used in many other real-time applications, some of them requiring dependable behavior. When designing CAN based dependable systems the bus topology of CAN is a single point of failure that needs to be addressed either through bus replication or by adopting alternative topologies with better fault-tolerance capabilities, e.g., star topologies. This paper proposes a set of components, an architecture and protocols to enable the dynamic management of the topology of CAN networks made of several replicated buses, both to increase the total available bandwidth and to reconfigure the network upon bus permanent error. In many operational scenarios, the proposed solution could be plugged into existing systems to improve its resilience to bus permanent errors, without changing the code running in the nodes.