Tag
ArcFlow 提出一种部署预算感知的训练算法，使单个流策略（flow policy）能在多个求解器步数预算下保持高性能，尤其在一步推理场景中优于五个基线方法。
提出Temporal Flow Policy (TFP)，将物理时间作为机器人策略记忆的缺失维度，使记忆状态随实测时间间隔演化，从而在扩散策略与π₀.₅上实现更准确的时序控制与记忆读出，在LIBERO及真实机器人任务中优于循环与状态空间记忆基线。
This paper formalizes multimodality in generative behavioral cloning and finds, through theory and extensive experiments on LIBERO, Meta-World, and synthetic benchmarks, that common simulation benchmarks are largely unimodal and that simple deterministic action transformers remain competitive with expressive generative policies such as diffusion, flow, and CVAE-based methods.
This paper introduces AALT, a composition-aware active imitation learning method that organizes demonstrations into latent topologies to maximize start-goal connectivity, improving task success in structured multi-task robotic domains with minimal expert queries.
Khushi shares her reading notes on Action Chunking Transformers and Diffusion Policy, explaining how action chunking with generative models like VAEs and diffusion improves imitation learning for robotics, and how they solve inference latency with decoupled planning and execution.
Push-Wiper is a robotics framework that reformulates viscous stain cleaning as an aggregation problem, using segmented pushing trajectories and a Diffusion Policy to generalize across stains, surfaces, and geometries, achieving up to 130% higher cleaning scores and zero-shot transfer.
DiPOD stabilizes diffusion policy optimization by interleaving self-distillation with policy-gradient updates to maintain a tight ELBO, preventing the double-drift phenomenon and achieving higher rewards in both language and continuous control tasks.
Introduces StereoPolicy, a framework that leverages synchronized stereo image pairs to improve geometric reasoning for robot manipulation policies, avoiding the fragility of RGB-D and point clouds. It integrates with diffusion-based and vision-language-action policies, showing consistent improvements in simulation and real-world tasks.
This paper introduces Parameterized Diffusion Policy (PDP), a framework that makes diffusion policies controllable by conditioning on low-dimensional latent parameters, enabling smooth behavior interpolation and adaptation without retraining. It demonstrates improved performance on complex multimodal robot tasks in simulation and real-world experiments.
Proposes Model-Based Diffusion Policy Optimization (MBDPO), a framework that unifies search and policy optimization in world models using diffusion policy representations, achieving consistent scaling behavior and superior performance across offline and online reinforcement learning tasks.
This paper introduces the Frequency Guidance Operator (FGO), a method for diffusion policies that smooths action generation by steering noisy samples through intermediate sub-frequency manifolds, improving performance on robotic manipulation tasks.