Agile methods for embedded systems

Abstract

The main goal to be answered by this Ph.D. thesis is whether there is a potential for a successful and powerful application of agile methods and related techniques to embedded systems development or not (cf. [2]). Regarding the special context of embedded system, there are some aspects to be mentioned as stated in [3]. These include the function oriented development which lead to early testing of the system, the use of target-processor simulation and the problem of hardware software codesign. The first problem being addressed is the evaluation of well known sub-techniques like refactoring, TDD, fast development cycles, short design horizon, or similar methods in the context of embedded systems. A complementary approach consists of the elaboration of underlying root causes which make agile methods appear as a sound alternative to classic techniques. For example assumptions like source code degrading over time, non-costumer oriented development, overly complex systems, and wrong development focus could be checked. A possible subdivision of the causes can be done by distinguishing effects that are generated by agile methods and effects that typically exist in embedded system engineering. This can be regarded as an alternative upside down procedure which will more likely yield a justification for agile methods in embedded system development. Finding a causation with an appropriate prioritization appears more challenging and thus will be used only to verify techniques which have a strong effect. The first step toward an assessment of agile methods has been started by executing a study during a lab course which is guided by the ideas described in [1]. Here, the students are divided into a planning group and an agile group, each developing a pre-crash system based on ultrasonic sensors. The two data collection mechanisms are a biweekly survey and a time recording log. The underlying aim is to show the influence of the planning horizon on embedded system development. This approach already sketches the main validation technique, which will be quantitative and composed of case studies and experiments. In addition, a case study with a high degree of variable control as proposed by [4] will be executed in order to guide the evaluation process to the most promising aspects of agile methods for embedded system development. © Springer-Verlag Berlin Heidelberg 2005.