It is important to find a collision-free path for an unmanned underwater vehicle (UUV) and manipulator, from an initial to a goal configuration, when considering automated vehicle activity in and around subsea structures. The problem is particularly acute when the combined motion of a vehicle and manipulator is considered, due to large numbers of degrees of freedom (DOF) which produce a large search space, the need for an efficient search algorithm, the need for defining cost functions without local minima, and an efficient representation of object geometries to avoid collisions. Over the past 20 years, a great deal of interest and progress has developed in robot path planning. This paper concentrates on efficient searching and object representation, while removing local minima. A novel approach to subsea vehicle/manipulator path planning using a nonlinear programming approach is presented. The central idea is to represent the free space of the workspace as a set of inequality constraints of a nonlinear programming problem, using the vehicle configuration variables. The goal configuration is designed as the unique global minimum point of the objective function. The initial configuration is treated as the start point for the nonlinear search. Then the numerical algorithm developed for solving the nonlinear programming problem is applied to solve the robot motion planning problem. Every immediate point generated using the nonlinear optimisation search method guarantees that it is in the free space and, therefore, is collision free. Mathematical foundations for constructive solid geometry, Boolean operations and approximation techniques are developed and are used to represent the free space of the robot workspace as a set of inequalities. The advantages of this approach are that mature techniques developed in the past thirty years, in nonlinear programming theory for the direction of search which guarantees convergence, efficiency and numerical robustness, can be applied directly to the robot path-planning problem. Simulation results show its effectiveness, efficiency and its potential as an online motion planner.