Category Theory-Based Synthesis of a Higher-Level Fusion Algorithm: An Example

Higher-level fusion (e.g., Level 2 that deals with derivation of relations among objects) often involves symbolic processing of information obtained from lower levels (e.g., Level 1, which deals with object detection, identification and tracking) which is based upon quantitative algorithms. The quantitative algorithms pass only some information to higher levels; some of the information is abstracted away in this process. However, this information might be needed by some of the rules at a higher level. Should in such a case the whole system be re-coded? In this paper we present an attempt to overcome the need for re-coding the Level 1 software. Instead, we propose to replace the activity of manually developing a fusion algorithm with an automated synthesis of the specification of the algorithm followed by automatic code generation. While solving such a problem in its entirety is a rather distant goal, in this paper we propose a solution to a more modest sub-problem. Rather than attempting to solve arbitrary information fusion problems, we assume that there exists a library of templates that specify information fusion objectives. The templates are formal specifications represented in a formal language. Since they are declarative, a variety of algorithms can satisfy their requirements. This paper presents an example of using a formal, category theory-based approach to the problem of synthesizing algorithms that satisfy templates for information fusion.