Towards Providing Debugging in the Domain-Specific Modeling - - PowerPoint PPT Presentation

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion

Towards Providing Debugging in the Domain-Specific Modeling Languages for Software Agents

Baris Tekin Tezel12 Geylani Kardas1

1Computer Science Department

Dokuz Eylul University Izmir, Turkey

2International Computer Institute

Ege University Izmir, Turkey

MDEbug 2018

1 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion

Plan

1

Introduction Multi-agent Systems Motivation Contribution

2

Debugging Approaches For MAS DSMLs Construction of a Mapping Between MAS Model Entities and the Generated Code Metamodel-based Description of Agent Operational Semantics

3

Discussion Evaluation of proposed approaches Alternative approaches

4

Conclusion Conclusion Future Work

2 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Multi-agent Systems Motivation Contribution

Multi-agent Systems

Software agents : are autonomous software entities acting to fulfill its duties on behalf of users. Multi-agent systems (MASs) : include multiple interacting software agents within an environment to provide solutions for complex systems which cannot be easily solved with individual agents or monolithic systems.

3 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Multi-agent Systems Motivation Contribution

What is the current position of MAS DSML?

Available MAS DSMLs usually support modeling both the static and the dynamic aspects of agent software from different MAS viewpoints. Although IDEs of these MAS DSMLs provide some sort of check and validation, they do not have a built-in support for debugging these MAS models. That deficiency causes the agent developers not to be sure

• n the correctness of the prepared MAS model at the

design phase.

4 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Multi-agent Systems Motivation Contribution

Why you need to do debugging to MASs on design phase?

MASs are distributed, concurrent, complex systems and as such are notoriously difficult to debug. Debugging abilities of most existing IDEs of MAS programming languages just rely on visualization of the messages passing between agents. Generally, software engineers use poor’s man debugger when they develop MASs. So, they need more powerful approaches against bugs in code.

5 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Multi-agent Systems Motivation Contribution

What are we planning?

This study aims at providing debugging inside MAS DSMLs. Thus, it is possible to complete the debugging phase at the modeling level before the code generation which leads to creating a MAS model conforming to the specifications at the beginning.

6 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Construction of a Mapping Between MAS Model Entities and the Gener Metamodel-based Description of Agent Operational Semantics

Construction of a Mapping Between MAS Model Entities and the Generated Code

7 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Construction of a Mapping Between MAS Model Entities and the Gener Metamodel-based Description of Agent Operational Semantics

Metamodel-based Description of Agent Operational Semantics

8 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Evaluation of proposed approaches Alternative approaches

Evaluation of proposed approaches

Pros of the first approach : Adopting existing, tried and well-known debugging facilities for the domain-specific models of software agents. Easy to implement. Cons of the first approach : The debugging is not performed at the same abstraction level (Semantic Gap). The approach assumes that all generated artifacts of a MAS DSML are executable.

9 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Evaluation of proposed approaches Alternative approaches

Evaluation of proposed approaches

Pros of the second approach : The implementation of this approach is independent from the GPL debuggers. The implementation would not need to consider situation of generated outputs (template code, non-executable etc.) Looks more suitable for complex modeling environments of MAS DSMLs. Cons of the second approach : Describing the runtime state of a MAS model to be added to the metamodel of the language. The writing of the rules of the M2M transformation which will refer to transitions between states.

10 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Evaluation of proposed approaches Alternative approaches

Alternative approaches

Creating intermediate instrumented version of the user-defined MAS model for supporting debugging activities. Providing xDSMLs for xDSMLs for MAS at first and then benefiting from the existing approaches on debugging xDSML.

11 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Conclusion Future Work

Conclusion

The application of the first approach is easier since it benefits from using already existing GPL debuggers. However, non-executable artifacts and generated template codes can make the application of the first approach inefficient. The second approach seems promising since it is free from underlying GPL structures. However, it is more difficult to apply because it needs addition of parts describing the runtime state of MAS model into the language metamodel and writing the corresponding M2M transformation rules.

12 / 13

Introduction Debugging Approaches For MAS DSMLs Discussion Conclusion Conclusion Future Work

Future Work

Future Work : Both providing implementation platforms for the proposed approaches and enriching them with alternative methods need further investigation which will be our future work.

13 / 13