1. Decision Support System for Fighter Pilots
The project started in November 2003 and ends in October It is carried out at DDRE in collaboration with the department of Operations Research (OR) at the Institute of Informatics and Mathematical Modelling (IMM) at the Technical University of Denmark (DTU).
Lars has a Master’s degree in Computer Science from the University of Copenhagen.
Lars Rosenberg Randleff
Ph.D. student
Danish Defence Research Establishment

2. Decision Support System for Fighter Pilots
Goals for the project
To make a model of the situation of the aircraft/pilot
To try different techniques for decision support based on the model
To make a prototype based on a given technique
In order to design a Decision Support System (DSS) for Fighter Pilots the first task would be to establish the parameters that have influence on the decisions made by the pilot, and to build a mathematical model based on these parameters.
The model depends partially on the technique used to make the DSS. The first attempt to make a DSS was done using a technique called Bayesian Network. (Bayesian Networks will be covered later on in this presentation.) Other techniques from Artificial Intelligence (AI) and Operations Research (OR) will be explored later on.
When some methods have been tried, and a “best solution” has been found, it will be time to develop a prototype of a DSS based on this method. The prototype should be able to give suggestion on actions to be taken based on recorded flight data from a PC-based flight simulator, or, preferably, from a data-bus on the aircraft itself.

3. Decision Support System for Fighter Pilots
Problem description
Many sources of information
Stress
Real-time
Trustworthy
The project faces some problems that must be taken care of, before a final DSS can be used. The first problem is to deal with the many sources of information. First of all it takes some labour to get date into the DSS, and second, the data from different sources may not always agree upon their view of the world outside the aircraft. Data from different sources should be fused before they can be used to make decisions from.
Any system to be used in the cockpit of a fighter aircraft should be designed for use under great stress. The DSS should have a high degree of user friendliness, and simply adopting design criteria from e.g. Windows software will not do. A pop-up window with the text: “The system has found a better solution than the one you are about to use. Are you sure you want to continue dispensing flares?” might be considered user friendly on a PC, while this is not the case in the cockpit.
To find the best decision for the pilot in a given situation might cause the DSS to calculate the outcome of a multitude of combinations of possible actions. These calculations should be done as fast as possible, since the DSS should not wait too long before suggesting the actions to the pilot. A delay of approximately 0.2 seconds is considered acceptable. This will be a challenge on most combinatorial problems.
Finally the pilot should have faith in the DSS, and it should not suggest actions that the pilot will disagree with. If it does the pilots may not use the DSS, and it will become a nuisance to him instead of a help.

4. Sources of information
Decision Support System for Fighter Pilots
Sources of information
Radar Warning Receiver (RWR)
Missile Warning System (MWS)
Jammer
Position and orientation
Amount of countermeasures
Chaff
Flares
Here are listed some of the sources that the DSS might use to make decisions from. The RWR, MWS, and jammer are used to model the threats in the world outside the aircraft. The pilot might see threats that is not seen by the sensors (e.g. an approaching missile not seen by the MWS), and he should react on these regardless of what the system tells him to do. It would be possible to design the system in such a way, that the pilot could inform it about threats seen by him, but this work would be cumbersome to the pilot, take to long to do, and the pilot would have lost any chance he had of reacting to the threat, by the time the system could tell him what to do.
Position and orientation is used to find e.g. aspect angles, and to decide on manoeuvres. The mode of the jammer, and the amount of remaining chaff and flares, is of course used to select the best countermeasure to use.

5. Decision Support System for Fighter Pilots
Missile guidance
Active Radar
Semi-active Radar
Infrared
Beam Rider
Anti-Radiation
Command
This slide is just to show some of the most used missile guidance systems in the world. For each threat the pilot (and hence the DSS) should consider which missiles that could be launched toward the aircraft, how they will be guided, and what actions he should do to avoid the missiles.
Track-Via-Missile
INS/GPS
Electro-Optical

6. Keeping it all together
Decision Support System for Fighter Pilots
Keeping it all together
According to the Missile Index about 650 missile systems have been developed
… and of these are still deployed
Different missiles, different types of guidance, different countermeasures
The Missile Index is a Japanese website ( that covers about 400 missile systems. It is based on the book “The Illustrated Encyclopaedia of World’s Missile Systems”. According to this website, about 650 different missile systems have been developed, and a third of these are still deployed somewhere in the world today. Even though neither the pilot nor the DSS will know the specifics of each of these systems, the number shows, that there is a lot of information for the pilot (and hence the DSS) to be aware of.

7. Decision Support System for Fighter Pilots
Data flow for the DSS
EWMS
DSS
Position data
Presentation
to the pilot
Situational Awareness
Mission data
Audio?
Display?
Force feedback?
Decision
Link-16 data
This figure shows a simple model of the data flow in the DSS. Data is fed to the DSS from the Electronic Warfare Management System (EWMS), and positional data as well as orientation data is taken from other onboard devices. Although this is not planned in the current project, mission data (route, mission, “hot spots”, etc) and Link-16 data from other aircrafts might be used by the DSS as well.
From the data fed to the DSS, the system calculates its own “Situational Awareness” (SA). This consists of a model of the current threats, their type, how long they have been there (and hence how far away an incoming missile might be), what countermeasures to take, and so on. From this model the DSS makes a short list of the best decisions on what to do.
The decisions are presented to the pilot in one way or another. Whether this should be done by audio, in a display, or by giving force feedback in the controls (or a combination of these) is not yet decided, and the decision is not within the scope of the project.

8. Decision Support System for Fighter Pilots
Pilot actions
Manoeuvre (altitude, aspect angle, G’s)
Dispense chaff
Dispense flares
Turn on/off jammer
Wait / do nothing
(Eject)
The aim of the DSS is to find the best action, or series of actions, in a given situation. For each combination of possible actions, a “survivability” for the pilot can be computed, and the combinations that gives the best survivability should be suggested to the pilot. These actions may include manoeuvring the aircraft to change its altitude or aspect angle to threats. The number of G’s, as well as the time it takes to make a manoeuvre, should be taken into consideration when suggesting actions to the pilot.
The system should also consider using countermeasures, or just wait to a threat becomes more threatening. Dispensing e.g. flares too early might have no effect, except on the amount of flares still on the aircraft.
Finally the system could suggest the pilot to eject himself from the aircraft. This is on this slide shown in parenthesises, as it should probably not be examined as one of the possible actions. In most cases “Eject” would yield the highest survivability, and the pilot could choose this, even if the only threat was a, possibly friendly, radar source that had been seen on the RWR.

9. Decision Support System for Fighter Pilots
A Bayesian Network
A Bayesian Network (BN) consists of a number of nodes (here shown as ellipses), and directed edges (arrows) between the nodes. Each node has a number of mutually exclusive states, and the node is in a given state with a given probability. The arrows show how the states of one node have influence on the states of another node.
When the states of one nodes change it will have influence on the nodes at the end of the arrows coming from the first node. Any changes in these nodes will then influence the following nodes, etc. This gives that a change in e.g. the “Missile” node (indicating whether or not a missile has been launched toward the aircraft) will have influence on the probabilities of the “Survive” and “Die” states of the “Survive” node at the bottom.
The rectangular nodes in the BN are action nodes. The DSS should find the combination of states within the action nodes that yields the highest probability of “Survive” in the bottom node. The green diamond nodes are utility nodes, and they are used to keep track of, how much chaff and flares that get dispensed.

10. Decision Support System for Fighter Pilots
Further work
Parameters are still added to the model
Other models and techniques are to be examined
Tests against simulator/aircraft flight data
Development of a prototype
Design of the interface between the DSS and the pilot
A lot of work still needs to be done before the final DSS can be implemented on an aircraft. First of all not all relevant parameters to be used in the model have been identified yet. New parameters might change the outcome of the DSS, and already added parameters might become irrelevant.
As mentioned earlier, other models and techniques need to be explored before a final approach can be selected. Stochastic Programming, Fuzzy Logic, and Artificial Neural Networks are some of the techniques that might be explored.
To test the developed models a framework for interfacing with flight data from either a flight simulator or a real aircraft needs to be implemented. The tests will not only help tuning the models for best performance, they will also show if the real-time demand has been met, and whether or not the system gives relevant and realistic suggestions. The development of a prototype will build upon the results of the tests.
Designing the interface between the DSS and the pilot is not within the scope of this project. It is an important issue though, since giving the pilot the best experience with the system is almost as important as to suggest to him the best solutions to a given situation.
Any questions to this presentation or to the project can be send by to Lars at