Change-Oriented Requirements Traceability.
Support for Evolution of Embedded Systems


Antje von Knethen
Fraunhofer Institute for Experimental Software Engineering*
Sauerwiesen 6, D-67661 Kaiserslautern, Germany
vknethen@iese.fhg.de



* The work was mainly performed at the Software Engineering Group, Department of Computer Science, University of Kaiserslautern,
Germany.
Abstract
Planning of requirements changes is often inaccurate
and implementation of changes is time consuming and
error prone. One reason for these problems is imprecise
and inefficient approaches to analyze the impact of
changes. This thesis proposes a precise and efficient
impact analysis approach that focuses on functional system
requirements changes of embedded control systems. It
consists of three parts: (1) a fine-grained conceptual trace
model, (2) process descriptions of how to establish traces
and how to analyze the impact of changes, and (3)
supporting tools. Empirical investigation shows that the
approach has a beneficial effect on the effectiveness and
efficiency of impact analyses and that it supports a more
consistent implementation of changes.


1 Introduction
Software systems have to be changed under various
circumstances during development and after delivery.
Modifications during development appear, for example, if
the requirements on the system themselves evolve or if a
system is developed incrementally. Modifications after
delivery of the system to correct faults, to improve
performance or other attributes, or to adapt the system to a
changing environment are called system evolution [1].
Impact analysis provides the entities that are
potentially affected in a system by a proposed change [2].
In our understanding, a system consists of code and
documents at higher abstraction levels, for example,
system requirements and system design. Traditionally,
impact analysis has been something that software
professionals do intuitively, after some cursory
examination of the system. This may be sufficient for
small systems, but it is not for large and critical ones. In
addition, empirical investigation shows that even
experienced software professionals predict incomplete sets
of change impacts [3].
The goal of our research was to develop a precise
impact analysis approach that improves planning as well
as implementing changes. A precise set of impacts can be
used by a project planner to accurately estimate the costs
of implementing a required change. Furthermore, a precise
set allows a maintainer to change a system consistently.
Inconsistencies within the system can be avoided and
ripple effects can be controlled. The changed system
includes fewer faults and is more cohesive.
Our impact analysis approach called TraceChange
consists of three parts: The first part is a fine-grained
conceptual trace model for embedded systems and a
product model [4] that is comparable to the one defined by
the Unified Process [5]. This trace model is the main
contribution of the thesis. The second part is a set of
process descriptions that determine how to establish traces
and to analyze the impact of changes. The third part is tool
support that provides (semi-) automatic impact analyses
and consistency checking of implemented changes [6].
Three empirical studies show the actual benefits of our
approach. All results strongly suggest that it has a
beneficial effect on the effectiveness and efficiency of
impact analysis and change implementation.
This paper gives a brief overview on TraceChange. It
is structured as follows. Section 2 describes deficiencies
of existing tracing approaches. Section 3 presents our
ideas of how to develop a trace model. Section 4
summarizes the process descriptions for setting and
analyzing relationships. Section 5 gives an overview on
the tool support. Section 6 describes our empirical
evaluation, and Section 7 points out future work.
Proceedings of the International Conference on Software Maintenance (ICSM’02)
0-7695-1819-2/02 $17.00 © 2002 IEEE