Software Architecture for Autonomous Vehicles

Abstract. Modern developments have increased the processing capacity of controllers along with the gradual decrease in their price. This has enabled the construction and use of intelligent control elements in embedded applications. These elements, such as smart sensors and actuators, not only can pre-process sensorial data using sophisticated algorithms but also enables the use of multiple communication protocols linking all the existing devices. The use of these control elements leads clearly to the construction of a distributed environment control in which the responsibility for processing environment data is divided between the several parts of the system. Moreover, there can be some redundancy in the data collected as well as the data collected by some sensors can have a greater priority level or must be collected using a shorter update deadline than others. In order to orchestrate this complex communication (which is natural on any distributed computing environment) between the elements of such a control system, the figure of a network connecting all the devices and a Middleware coordinator are evident. This article firstly analyzes the current efforts in developing the software architecture used to support these requirements in terms of message coordination and processing. Based on these, a software architecture is then proposed in order to enable the easy integration of new control elements dynamically as well as maintaining the real time response of the control system