Multi-robot systems can be useful in applications such as map building, surveillance, and search and rescue. To be efficient in practice, the robotic team should cooperatively explore a region of interest. However, when the environment is unknown, it is challenging to plan collision-free paths in real-time under both non-holonomic mobility constraints and energy constraints. We propose the Multi-robot Hex Decomposition Exploration (M-HDE) method for multiple Dubins vehicles to explore unknown environments. M-HDE ensures robots return to the initial (departure) position before energy runs out. Furthermore, when available energy level allow, M-HDE can achieve complete exploration of the environment. The proposed approach generates smooth, continuous paths for Dubins vehicles to follow at constant velocities, and offers geometric closed-form solutions for both team formation and paths. The performance is evaluated with a team of Turtlebots in simulated environments with obstacles in Gazebo.