Search papers, labs, and topics across Lattice.
The paper introduces iDMa, a stochastic trajectory prediction framework that combines diffusion models with the Mamba architecture for improved accuracy and efficiency. iDMa employs a dual-parameter learning mechanism to optimize noise estimation in both mean and variance spaces, along with a hybrid denoising backbone network using Transformer encoders and Mamba blocks. Experiments on ETH-UCY and SDD datasets demonstrate that iDMa achieves state-of-the-art performance, reducing average displacement error (ADE) compared to existing methods.
Diffusion models get a Mamba-powered upgrade, yielding state-of-the-art results on pedestrian trajectory prediction.
Trajectory prediction constitutes a key technology for intelligent systems to forecast future movements of dynamic agents, yet it faces significant challenges due to the uncertainty of motion behavior. We propose iDMa, a stochastic trajectory prediction framework that pioneers the integration of diffusion model with Mamba architecture to achieve high-precision and high-efficiency trajectory generation. Our approach introduces two key innovations: (1) a dual-parameter learning mechanism that optimizes noise estimation of mean and variance space, unlike conventional diffusion methods that employ fixed variance during the denoising process, so as to constrain the feasible domain more accurately; (2) a hybrid denoising backbone network that incorporates Transformer encoders and Mamba blocks. Compared to existing state-of-the-art methods, iDMa reduces the average displacement error (ADE) by 4.76% (0.20 vs. 0.21) on the ETH-UCY dataset and 1.85% (7.95 vs. 8.10) on the SDD dataset.