In developing a traffic simulation in which each vessel has a degree of on-board "intelligence" it soon becomes clear that there are shiphandling situations for which the correct course of action is by no means apparent. These dilemmas no doubt mirror similar situations in the real world, and seem to occur with the following scenarios:

　

　

 Not only do these dilemmas arise, there are other uncertainties when developing a rule-based dynamic simulation. Among these are:

　

　

 Not all of these issues have been solved in the software, but some progress has been made. This section discusses what has been done and explores it by means of an example.

　

　

　

 The dilemma in this case arises when a vessel has to make an avoidance manoeuvre in an area where it would otherwise need to make a change of course in normal navigation. This might occur where a route has a curve or a course change is required when rounding a piece of land. Should priority be given in all cases to the avoidance manoeuvre, even if it were to result in the ship moving off course, possibly into further danger?

　

 In a real situation, a judgement would presumably be made by weighing up the amount of risk in each of the alternatives. A simpler approach is adopted in the program. Two candidate rudder angles are computed, one for normal navigation, the other for the avoidance manoeuvre. The size of each angle is taken as a measure of the importance of the manoeuvre. In all cases, if that for the avoidance manoeuvre is the largest, it takes precedence and is chosen by the program. If the angle for navigation is the larger, it is subject to two further tests. If the avoidance manoeuvre is over, then clearly the "navigation" angle is used. If it is not, then the angle for navigation only assumes precedence if that for avoidance is sufficiently low, say 2°. In this way, a judgement is made, weighted towards the avoidance manoeuvre, on the assumption that the normal course can be resumed once the more important avoidance manoeuvre is over.

　

　

 No special arrangements are made for multiple encounters; all ships manoeuvre according to the steering rules in force.

 It is assumed that two vessels are approaching one another at 90° on a collision course. This is shown in Figure 2. The vessels are assumed to obey the ColRegs meaning that ship A is the stand-on vessel and ship B the give-way. This simple encounter is discussed first, and then the multiple encounter is introduced. In this, Ship B must still alter course (to starboard), but the situation is complicated by ship C on her starboard side, which she is overtaking. A further complication arises because ship A is also overtaking a vessel on her starboard side (ship D). All four ships start the run in positions which would result in them being at the crossing point simultaneously if no course or speed changes were made.

　

　

 Both scenarios was set up on Dymitri using the ColRegs steering rules in an attempt to determine how the program would deal with such situations. It was assumed that ships A and B were moving at 20 knots and were of container ship size. Ships C and D were assumed to be smaller general cargo vessels moving at 15 knots.

　

 The results are shown in Figures 3 to 5. Figure 3 shows the initial case of ships A and B only. Both are on a collision course and ship B is seen to give way and pass under the stern of ship A. This avoidance manoeuvre is in accord with the ColRegs and, as well as a change of heading, it involved a reduction of speed by ship B of some 3.5 knots. It is important to emphasise that this reduction in speed helped considerably towards the success of the avoidance manoeuvre.

　

　

 However, the change of course was quite violent and involved hard-over rudder angles. This is because the minimum allowable CPA was set at 200 metres, somewhat more than the half-length of ship A. Reducing this value to, say, 50 metres, produces the track in Figure 4. The avoiding action was considerably deduced, but a near miss, or possibly a glancing collision would have resulted. Slowing down more, or beginning the avoidance manoeuvre earlier, would have helped, but it is apparent that in the simple two ship encounter, the simulation model can produce reasonably realistic representation s of real situations.

　

 Tracks from the multiple encounter are shown in Figure 5. This is interesting because ships A and B elected to overtake ships D and C respectively by altering course to starboard at the beginning of the run. There was enough room to do this while the overtaken ships (C and D) maintained speed and heading. Ship B also reduced speed as it approached the collision zone.