Tag
The article presents BFN-RL, a unified generative modeling framework for offline reinforcement learning based on Bayesian Flow Networks, capable of generating effective trajectories across discrete and continuous state spaces.
This article explains how to use polynomials, specifically smoothstep functions, to create smooth motion trajectories for robots and 3D printers by matching derivatives of position (velocity, acceleration, jerk, etc.) to avoid discontinuities.
This paper introduces a permutation-equivariant neural operator that maps spacecraft, target, and debris distributions to collision-aware trajectories for entire swarms in one forward pass, trained with self-supervised physics objectives and adversarial threats. It generalizes zero-shot from 10 to 1,000 spacecraft amid dense debris, matching optimal-control accuracy while reducing proximity.
Fast-dDrive is a block-diffusion VLA model for end-to-end autonomous driving that achieves state-of-the-art trajectory accuracy while delivering over 12x throughput speedup over autoregressive baselines, addressing the trade-off between high-fidelity planning and efficient inference for edge deployment.
ReflectDrive-2 is a new discrete diffusion planner for autonomous driving that uses reinforcement learning to enable self-editing of trajectory tokens, achieving high performance and low latency on the NAVSIM benchmark.