The problem of computing the euclidean shortest path between two points in the three dimensional space bounded by a collection of convex disjoint polyhedral obstacles is known to be NP-hard and in exponential time for arbitrarily many obstacles. It can be solved in O(n²) time for single convex polyhedron obstacle (here n is the total number of vertices of polyhedron). In this paper, the author mainly researchs the shortest problem of the case of two convex polyhedral obstacles, and presents an algorithm that solves this problem in O(n²) time, and improves improving significantly previous best result On³logn) for this special case. On the other hand, the author also presents a better result O(∑12^i-1n²_i) for the problem of shortest path amidst a fixed number of convex polyhedral obstacles.