In this paper, an optimization method for a redundant serial robotic manipulator's structure is proposed in order to improve their performance. Optimization was considered in terms of kinematics using the proposed objective function and the non-linear Levenberg-Marquardt algorithm for multi-variate optimization. Range limits of the joints, bounds of the design parameters, and a constrained workspace are enforced in the proposed method. A desired manipulator can be optimized to cover the required task points using dimensional synthesis. This approach effectively optimizes the link lengths of the manipulator and minimizes the position and orientation errors of the tool center point. A commercial heavy-duty hydraulic, underground tunneling manipulator was used to demonstrate the capability of the proposed optimization method. The obtained results encourage the use of the proposed optimization method in automated construction applications, such as underground tunneling, where the confined environment and the required task add challenges in the design of task-based robotic manipulators.