在太空环境下实现自适应抓取，通过潜空间学习提升样本效率与鲁棒性。
Towards Space-Based Environmentally-Adaptive Grasping
- 在融合多模态的潜空间中学习控制策略，结构化决策过程。
- 不到100万步即达95%以上成功率，收敛速度优于主流视觉基线。
- 适合研究机器人自适应抓取、强化学习在极端环境中的应用。
在非结构化环境中进行机器人操作需要在多样条件下可靠执行，但当前许多先进系统仍受限于高维动作空间、稀疏奖励和训练后泛化能力弱的问题。本文以太空环境下的抓取任务为例，研究这些局限。我们直接在融合多模态信息的显式潜流形上学习控制策略，实现结构化决策。基于GPU加速的物理仿真，构建单次执行的抓取任务，在连续变化的抓取条件下，采用软演员-评论家（SAC）强化学习算法，在不足100万环境步内实现超过95%的任务成功率。实验证明，该方法在相同开环单次执行条件下，收敛速度优于代表性视觉基线。分析表明，显式在潜空间推理显著提升了样本效率，并增强了对新物体形状、夹持器几何、环境杂乱度和传感器配置的鲁棒性。最后识别了现存挑战，并指明了迈向完全自适应、可泛化的太空抓取系统的方向。
原文摘要 · Abstract (English)
Robotic manipulation in unstructured environments requires reliable execution under diverse conditions, yet many state-of-the-art systems still struggle with high-dimensional action spaces, sparse rewards, and slow generalization beyond carefully curated training scenarios. We study these limitations through the example of grasping in space environments. We learn control policies directly in a learned latent manifold that fuses (grammarizes) multiple modalities into a structured representation for policy decision-making. Building on GPU-accelerated physics simulation, we instantiate a set of single-shot manipulation tasks and achieve over 95% task success with Soft Actor-Critic (SAC)-based reinforcement learning in less than 1M environment steps, under continuously varying grasping conditions from step 1. This empirically shows faster convergence than representative state-of-the-art visual baselines under the same open-loop single-shot conditions. Our analysis indicates that explicitly reasoning in latent space yields more sample-efficient learning and improved robustness to novel object and gripper geometries, environmental clutter, and sensor configurations compared to standard baselines. We identify remaining limitations and outline directions toward fully adaptive and generalizable grasping in the extreme conditions of space.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。