Abstract: To address the complex path planning requirements and improve path smoothness for sorting robots operating under extreme temperatures, this paper proposes a 3D path planning method based on 5G communication. Firstly, a 3D spatial environment model is established for extreme temperature conditions to simulate the distribution of obstacles and temperature variations under low, normal and high temperatures, ensuring the authenticity and adaptability of the environment model. Secondly, 5G communication technology is adopted to transmit 3D spatial signals and position data of robot coordinate points to the network in real time. The Next-Best-View (NBV) algorithm based on deep reinforcement learning, together with the criteria of information gain and consumption cost, is used to optimize the observation perspective and moving route of the robot, so as to achieve smoother and more efficient paths. Simulation results show that the proposed method can generate shorter and smoother motion paths in extreme temperature environments. Practical application verifies that it delivers excellent path planning performance and strong robustness in complex scenarios, which provides reliable technical support for the operation of sorting robots under extreme temperature conditions.