We now turn to the Basic Variable-Endpoint Control Problem. In this case there is an additional statement to be proved, which is the transversality condition (4.3). In Section 4.2.6 we had that the terminal cone was separated from the ray ; the reason for this was that hitting a point on below contradicted optimality. When the fixed endpoint is replaced by the surface , we would instead have a contradiction with optimality if we were able to hit a point with a cost lower than whose -component is in (but is not necessarily ). Let us denote the set of such points by . We are looking to establish separation between convex sets; for this reason, just as we replace the actual set of terminal points with its linear approximation , we will consider the linear approximation of given by the linear span of and the tangent space , i.e., the set
This lemma is proved by an appropriate generalization of the argument we used to prove Lemma 4.1. Suppose that the statement is not true. Then we can find a point in which is contained in together with some -ball around it. We can write this point as
for suitable and (moving slightly down if necessary, we can ensure that it does not lie on the upper boundary of ). Since belongs to , each point in is given by where is a first-order terminal state perturbation arising from a temporal and/or spatial control perturbation. We know that the corresponding exact terminal states are of the form and form a ``warped" version of , which we denote by .
This construction remains valid as . The ball is centered at , its radius is , and the ``warping" that produces is of order . As in Exercise 4.5, it can be shown that contains a ball centered at whose radius is of order . Furthermore, since and are tangent to each other along , the distance from to is also of order . Hence, for small enough, actually intersects . But this, as we already noted, contradicts optimality of , and the lemma is established. The preceding argument is illustrated in Figure 4.13, where the plane and the curved surface represent and , respectively, the shaded object is the portion of that lies between and , and the ray in containing , is also shown.
By Lemma 4.2 and the Separating Hyperplane Theorem, there exists a hyperplane that separates and . We denote its normal vector by (4.28) as before. In view of the definition (4.37) of and the fact that belongs to , the separation property still gives us the inequalities (4.29) and (4.30). Thus all the constructions and conclusions of Sections 4.2.8 and 4.2.9 still apply, and so we know that the first three statements of the maximum principle are true. On the other hand, writing the separation property for vectors in with (the -component) equal to 0, we obtain the additional inequality
Note that in the special case when (a free-time, free-endpoint problem), the hyperplane must separate from the entire -dimensional half-space that lies below . Clearly, this hyperplane must be horizontal, hence its normal must be vertical and we conclude that . This is consistent with (4.3) because in this case.