The fourth Industrial revolution has put robots at a starring position in every endeavor. For its correct operation, testing is a crucial phase before deployment. Running tests on robots is more challenging than testing on software, particularly when accounting for replicability, environment matching, recovery strategies, etc. In the case of robots for marine environments, running tests is even more challenging because of the features of the former such as high dynamism, poor visibility, extreme conditions, etc. New technologies such as AUVs (Aquatic Unmanned Vehicles) have gained attention since they are reliable, affordable, and highly maneuverable. Even for these platforms, sufficient trials should be conducted and several outcomes studied before deployment, to make it secure and more likely to accomplish the objectives. We propose a strategy that provides the capabilities of running trials in simulated environments. Once the trials are conducted and a plan is determined, the real-world deployment of the robot will be more reliable and more likely to achieve its goals. Our strategy is based on the RRT algorithm (Rapidly exploring Random Trees) for an AUV. We present a variation that considers kinematics, dynamics, and uncertainty of the vehicles' movement, allowing safe experimentation for the robot, environment and researchers.