This paper introduces LDP, a local diffusion planner for efficient robot navigation and collision avoidance. The method addresses the challenges of data urgency and myopic observation in robot navigation by collecting expert policy data from diverse scenarios and incorporating global paths as additional conditions for the diffusion model. The LDP algorithm is trained using a dataset of expert policies with two different preferences, and it is enhanced by incorporating global observations in a lightweight manner. This enhancement broadens the observational scope of LDP, effectively mitigating the risk of becoming ensnared in local optima and promoting more robust navigational decisions. Experimental results demonstrate that LDP outperforms other baseline algorithms in navigation performance, exhibiting enhanced robustness across diverse scenarios with different policy preferences and superior generalization capabilities for unseen scenarios. The LDP algorithm is also deployed on real-world robotic platforms, showcasing its competitive advantages. The main contributions of this work include the introduction of LDP, a dataset of expert policy based on 2D laser sensing, the use of global paths as additional guiding conditions, and extensive experiments demonstrating that LDP outperforms other baseline algorithms in terms of superior navigation performance, stronger robustness, and more profound generalization capabilities. The paper also highlights the effectiveness of the algorithm by deploying it on physical robots, thus highlighting its practical value.This paper introduces LDP, a local diffusion planner for efficient robot navigation and collision avoidance. The method addresses the challenges of data urgency and myopic observation in robot navigation by collecting expert policy data from diverse scenarios and incorporating global paths as additional conditions for the diffusion model. The LDP algorithm is trained using a dataset of expert policies with two different preferences, and it is enhanced by incorporating global observations in a lightweight manner. This enhancement broadens the observational scope of LDP, effectively mitigating the risk of becoming ensnared in local optima and promoting more robust navigational decisions. Experimental results demonstrate that LDP outperforms other baseline algorithms in navigation performance, exhibiting enhanced robustness across diverse scenarios with different policy preferences and superior generalization capabilities for unseen scenarios. The LDP algorithm is also deployed on real-world robotic platforms, showcasing its competitive advantages. The main contributions of this work include the introduction of LDP, a dataset of expert policy based on 2D laser sensing, the use of global paths as additional guiding conditions, and extensive experiments demonstrating that LDP outperforms other baseline algorithms in terms of superior navigation performance, stronger robustness, and more profound generalization capabilities. The paper also highlights the effectiveness of the algorithm by deploying it on physical robots, thus highlighting its practical value.