ADM - Autonomous Decision Maker (PIITSA)
The main goal of this project is to define alogical design of ADM that is general, well structured and easy to use in order to be of practical value. A goal is also that this work can lead forward to the creation of a physical demonstrator.
The background to this project is that when control shall be made using the Internet, and when the application is safety critical, it is not suitable to let the whole responsibility be handled by the application. A separate and possibly standardized unit is instead preferred. The intention of ADM is that it should be an intelligent and autonomous interface function to the Internet with focus on safety and real-time properties.
The work is divided into the following activities:
Activity 1 - Survey of existing algorithms and methods
This activity indludes a survey of applicable algorithms and methods that could be applicable and their properties. The purpose is to get a picture of ”state-of-the-art” concerningresearch applications within the domain. A result is also specify if new algorithms and methods are needed.
Activity 2 - Specification
This activity includes formulation of requirements for ADM. A fundamental task is to investigate how safety aspects can be mapped to/from application and from/to ADM. One approach is to specify a langugae that defines what is possible to express. Flexibility is achieved e.g. by requirements on configuration. A special aspects is if it possible extract knowledge from one ADM and use it to learn another.
Activity 3 – Integration
This activity includes a study of how to integrate the parts described in activity 1 and 2 into a cooperating unit. Here is also included the execution model. The result of this actvity is the creation of a logical demonstrator.
Activity 4 – Evaluation
This activity includes an evaluation of the ADM concept and the logical demonstrator