NIST Assembly Benchmark

该数据集包含了针对不同工业装配任务的深度强化学习（DRL）策略性能评估的实验试验，重点关注这些策略在不同条件下的鲁棒性和可靠性。数据集涵盖了在各种条件下的试验，包括扰动误差和动态环境，旨在评估DRL策略的泛化能力。该数据集规模为13096次试验，成功率达到了99.8%。涉及的工业装配任务包括电路板插入、HDMI插入和钥匙锁插入等。

This dataset comprises experimental trials for evaluating the performance of deep reinforcement learning (DRL) policies across diverse industrial assembly tasks, with a primary focus on the robustness and reliability of these policies under varying conditions. The dataset includes trials conducted under various scenarios including perturbation errors and dynamic environments, aiming to assess the generalization capability of DRL policies. The dataset totals 13,096 trials, with a success rate of 99.8%. The involved industrial assembly tasks include circuit board insertion, HDMI insertion, key-lock insertion and other similar tasks.

搜集汇总

背景与挑战

背景概述

NIST Assembly Benchmark是一个包含工业机器人装配任务视频的数据集，涵盖多种连接器类型如USB、DSUB和防水连接器的插入操作，以及移动目标场景。这些视频用于支持多模态策略学习研究，但数据集详情页面未提供额外元数据或任务说明。