This research project proposal focuses on very important subjects that are mentioned in the strategic plan for software engineering developed by the European Institute of Software (EIS). In this plan, it is stated the necessity of elaborating abstract models for the activities of modeling, implementation, and maintenance of software systems and it is proposed to automate the processes related to these activities in order to increase the competitiveness in the software engineering domain. The project's main goal is to assure fundamental knowledge development in a PNCDI-II essential domain 1.3. Mathematics, approaching two major research directions: 1.3.1 Theoretical informatics and 1.3.5 Mathematical models.

Objectives After an in-depth study of the methods, techniques and results reflected by the state of the art of the fields of maintenance and evolution of software systems (Section 7.1), we have concluded that applying machine learning techniques and the agent based technology would lead to valuable research results in the direction of the adaptive optimization of software systems and will open new and very promissing research directions.

The main objective of the project is to use artificial intelligence techniques (machine learning) for structural adaptation of software systems during their maintenance and evolution and for their behavioral self-adaptation, as well. We aim at developing abstract formal models for the above mentioned problems in order to assure their precise modeling and to create valid conceptual models. We also intend to develop an intelligent multiagent decision support system for assisting software developers in the maintenance and evolution of software systems. The importance of the proposed project with repect to the objectives of the national strategy for research, development and innovation is emphasized by the following scientific objectives:

• O1. The elaboration of new machine learning based methodologies for structural and behavioral adaptive optimization of software systems.
• O2. The theoretical validation of the above mentioned methodologies, by elaborating some formal methods for problems related to program comprehension: design patterns identification and concept location.
• O3. The elaboration of theoretical models for software reingeneering problems: refactoring and introducing design patterns in existing software systems.
• O4. The development of machine learning algorithms (clustering, self-organizing maps, neural networks) for the activities related to program comprehension and structural modifications (optimizations) of software systems (design patterns identification, concept location, aspect mining, refactoring, and introducing design patterns).
• O5. The development of machine learning algorithms (clustering, self-organizing maps, neural networks) for behavioral self-adaptation of software systems (for optimizing the communication with other systems - access to resources, services, depending on the activities workflow from the system-, for self-adaptation of the data structures used depending on the data workflow from the system).
• O6. The conceptual modeling, using the agent based technology, of a system to allow adaptive optimization (both structural and behavioral) of software systems. This model will be based on the theoretical models previously developed.
• O7. The incremental development, based on the previous conceptual model, of an intelligent multiagent decision support system for assisting software developers in the maintenance and evolution of software systems, named DSSEM (Decision Support System for software Evolution and Maintenance).
• O8. The study of improving the performance of DSSEM multiagent system by ysing parallel programming.
• O9. The experimental validation of DSSEM system on real software systems.
• O10 The Publications of the theoretical and applicative scientific results of the project, both at national and international level.

The activities carried on during the research project are essentially fundamental research activities in the domain of information and communications technology. Activities from the category • fundamental basic researches in information technology and applied mathematics are used for accomplishing the objectives O1, O2, O3, O4, O5, O8. Activities from the category applicative researches in computer science are used for accomplishing the objective O6. Activities from the category • analysis, design, programming and implementation of complex software systems are used for accomplishing the objectives O7 and O9. Activities from the category • Publications of the results of researches in computer science are used for accomplishing all the objectives (O1-O10).

We also aim at creating an interdisciplinary research group where the young researchers will have the opportunity of obtaining research results competitive at international level. We intend to continue the collaboration with the SEVERE (Software Vizualization and Evolution Research Group) research group from the Computer Science Department of “Wayne State” University from Detroit. This collaboration was already started and materialized in some publications ([SER06], [CAM06]) with the vicemanager Assoc. Prof. Andrian Marcus (http://www.cs.wayne.edu/~amarcus/). We also intend to initiate a collaboration with the researchers from the Contract project (Bonn University, Germany), in the direction of formalizing refactorings.

Original elements The proposed project aims to bring original contributions on the theoretical background and on the existing methodological support for the analysis of the problems related to the maintenance and evolution of software systems and to add original theories and methodologies also. The original elements that the present project is focused on can be found in the previously proposed objectives, especially in objectives O1-O7. The original results on the project topic obtained until now collectively by the team members (as are described in Section 8.2) confirm that the proposed research directions are very promising, and will provide valuable scientific results. Considering these aspects, we estimate that the original elements aimed by the present project will have a large impact and a great importance in developing the scientific knowledge of the approached domains.

The estimated impact of the present project on the scientific knowledge: Impact on the fundamental research. On the one hand, the project is focused on the theoretical fundamentals of the problems of adaptive optimization of software systems using machine learning techniques and multiagent systems. As we are focusing on the development of the conceptual and theoretical background, our researches will have a remarkable impact on the theoretical research of the mentioned domains. Impact on the applicative research. On the other hand, the problems approached by the project are derived from practical requirements of software engineering. Consequently, we will use the scientific knowledge in order to develop the multiagent system DSSEM. Economical impact. The multiagent system DSSEM can be used in the development of real software systems and can make the maintenance and evolution of software systems more efficient.

Interdisciplinary character. By the proposed objectives, considering the original elements focused upon, the present project has a strong interdisciplinary cahractetistic and will certainly play an important role in the knowledge development from the fields of information technology - artificial intelligence and applied mathematics.