This paper introduces Equivariant Diffusion Policy, a novel method for robotic manipulation that leverages domain symmetries to improve sample efficiency and generalization in the denoising function. The method is theoretically analyzed for SO(2) symmetry in 6-DoF control and evaluated on 12 simulation tasks from MimicGen, showing an average success rate 21.9% higher than the baseline Diffusion Policy. The method is also tested on a real-world system, demonstrating effective policy learning with fewer training samples compared to the baseline. The key contributions include the proposal of Equivariant Diffusion Policy, theoretical analysis of equivariance, and empirical evaluation in both simulated and real-world environments.This paper introduces Equivariant Diffusion Policy, a novel method for robotic manipulation that leverages domain symmetries to improve sample efficiency and generalization in the denoising function. The method is theoretically analyzed for SO(2) symmetry in 6-DoF control and evaluated on 12 simulation tasks from MimicGen, showing an average success rate 21.9% higher than the baseline Diffusion Policy. The method is also tested on a real-world system, demonstrating effective policy learning with fewer training samples compared to the baseline. The key contributions include the proposal of Equivariant Diffusion Policy, theoretical analysis of equivariance, and empirical evaluation in both simulated and real-world environments.