Recently, Neural Radiance Fields (NeRF) have become mainstream in robotic mapping since they represent the geometric features of a scene well using only a fully connected MLP with low memory requirements and without the need for an expensive vision sensor. In this paper, we present a novel approach for real-time motion planning of robots with different heights based on deep reinforcement learning and NeRF. We leverage NeRF to construct detailed maps by encoding volume densities at various heights, facilitating accurate environmental representation. By segmenting the 3D scene according to specified robot heights, a 2D map is generated with the occupancy information pertinent to each height. A deep reinforcement learning(DRL) framework incorporating NeRF-generated volume densities into the reward function is designed to enhance the agent’s ability to navigate effectively. The reward function minimizes collision probability and encourages optimal path length, with volume density as a critical collision avoidance component. Experimental results demonstrate the proposed DRL framework can plan the motion of mobile robots with different heights in real time.