In specifying motion toward a gap, the system must initially aim to position the moving object at a point from which it can negotiate the first obstacle to a straight path through the gap. This point is usually next to the junction at the end of an obstructing segment. Later motions can then be expressed in terms of further obstruction-passing points. These sub-goals for motion past obstacles are called ``via points''. The currently implemented system only considers via-points whch can be reached using straight line motion, but a more general system could include motion along curved lines, and also include information regarding the required rotational orientation of the moving object.
The system locates via points by finding what obstacle is on the near side of a suitable gap, and then searching for the first junction at the entrance to the gap. The via point is specified as a pair of boundary elements: this junction, and the nearest boundary element on the other side of the gap.
The proposed motion is specified as the direction from one boundary element on the moving object to one of the via point elements. This direction is then translated into a direction of motion for each boundary element of the moving object, so that any further obstructions to motion may be detected.
Directions of motion are also evaluated as to how ``vague'' they are. This is discussed further in the next section; it provides some robustness by allowing the system to evaluate the quality of its directional information, and employ precise directions rather than vague ones where possible.
Motion specification is currently only translational - if rotational motion was to be accomodated, substantial changes would be required. The simplest way to extend the translational planning system would be to use sequential combinations of rotational and translational motion. Reasoning about simultaneous rotation and translation in two dimensions would require the use of a 3-dimensional configuration space, and new planning strategies and directional reasoning, in addition to new methods for representing gap size, finding moving object extremities, and specifying via points.