This work presents a novel path planner, DE3D-NURBS, for 3D path planning considering the maximum and minimum climb/dive angle and the maximum curvature imposed by a robot, such as an autonomous fixed-wing aircraft. To deal with this constrained problem, we propose a Differential Evolution (DE) based algorithm, assuming the partitioning of feasible and infeasible individuals and a novel mutation operator. This solution can provide a short differentiable smooth path, directly parameterized by a kth-order Non-Uniform Rational B-Spline (NURBS) curve, a generalization of B-spline curves. This representation provides more flexibility to satisfy essential kinematic constraints and to take into account obstacles and topography. This path can guide an Unmanned Aerial Vehicle (UAV) through a set of desired robot configurations (position and orientation) in 3D space. The simulation results provide a more comprehensive insight into the capabilities of the proposed planner to generate feasible paths, considering simple scenarios in a free obstacle environment and also complex scenarios. Such scenarios include topographic environments, narrow passages such as a tunnel, and a solution for closed-loop missions in a modular approach. We also provide a comparative benchmark with other nature-inspired algorithms. These results serve as a baseline for evaluating the effectiveness of the proposed DE algorithm in complex and challenging case studies.