TDF

We have developed and continue to build on a methodology and accompanying tool, TDF (Tactics Development Framework), based on the BDI (Beliefs, Desires, Intentions) paradigm. TDF supports structural modelling of missions, goals, scenarios, input/output, messaging and procedures, and generates skeleton code that reflects the overall design. TDF has been evaluated through comparison with UML, indicating that it provides significant benefits to those building autonomous, tactical decision-making systems.

There is an increasing need for autonomous systems that exhibit effective decision-making in unpredictable environments. However, the design of autonomous decision-making systems presents considerable challenges, particularly when they have to achieve their goals within a dynamic context.

Tactics designed to handle unexpected environmental change, or attack by an adversary, must balance the need for reactivity with that of remaining focused on the system’s overall goal. The lack of a design methodology and supporting tools for representing tactics makes them difficult to understand, maintain and reuse. This is a significant problem in the design of tactical decision-making systems.

Although our main user base is in the defence industry, the methodology and tools we develop are applicable to any simulation or autonomous system that involves modelling human behaviours and strategies such as bushfire simulations, UAVs and disaster management systems.

The following video provides an introduction to TDF:

Diagrammatically Model Strategy and Tactics

A strategy is a general approach to a problem, whereas a tactic is a specific method of achieving an objective and often needs to be adaptable to handle unexpected changes in the situation faced. A submarine commander’s use of stealth to approach a target is a strategy, whereas the particular method used, for example hiding in the adversary’s baffles (a blind spot), is a tactic. Tactics are concerned with the current, unfolding situation – that is, how to deflect threats and exploit opportunities to achieve one’s objective. TDF offers an intuitive, diagrammatic means of modelling tactics, strategy and dynamic decision making in general.

Engineer Agent and Team-Oriented Systems

Agent-Oriented Software Engineering (AOSE) addresses the problems of specifying, designing, validating, maintaining and deploying agent-based systems. As a direct descendant of the Prometheus AOSE methodology, TDF’s lineage spans some 20 years. TDF extends Prometheus with richer goal diagrams, team modelling, tactics and plan diagrams.

Build Explainable AI

Increasingly, artificial agents and humans need to work together in teams. To function effectively as a team, there has to be mutual understanding and trust between team members; the human must comprehend what the artificial team members are doing and why they are behaving as they are. Thus, the agent must expose its reasoning to the human, and one way to do this is to leverage TDF’s intuitive, diagrammatic representation so that the human can see what goals the artificial agents are pursuing and how they are trying to achieve their goals.

Elicit Knowledge from Domain Experts

In general, before one can model decision-making, one has to elicit the details from human experts. This turns out to be a very challenging endeavour because much of human expertise is tacit, and so it is difficult for domain experts to introspect about why they make the decisions they do. This knowledge acquisition bottleneck was recognised very early on as a key problem in the construction of intelligent systems. TDF supports the elicitation of tactical expertise by focusing on how decision making in dynamic environments requires one to notice change, determine whether the current course of action is still effective, and if necessary choose an alternative approach.

TDF Tool and Book

The TDF Tool supports the creation of behaviour models. It is a free-to-use standalone application that runs on Linux, macOS and Microsoft Windows. It is currently being reimplemented and so is not available for download at this time.

The TDF book on modelling dynamic decision making has now been published. You can find out more here