This thesis describes a technique to obtain an optimal collision-free path for an automated guided vehicle (AGV) and/or robot in two and three dimensions by synthesizing a B-spline curve under geometric and intrinsic constraints. The problem is formulated as a combinatorial optimization problem and solved by using simulated annealing. A two-link planar manipulator is included to show that the B-spline curve can also be synthesized by adding kinematic characteristics of the robot. A cost function, which includes obstacle proximity, excessive arc length, uneven parametric distribution and, possibly, link proximity costs, is developed for the simulated annealing algorithm. Three possible cases for the orientation of the moving object are explored: (a) fixed orientation, (b) orientation as another independent variable, and (c) orientation given by the slope of the curve. To demonstrate the robustness of the technique, several examples are presented. Objects are modeled as ellipsoid type shapes. The procedure to obtain the describing parameters of the ellipsoid is also presented.