智能船舶远程驾驶负荷在线监测数据集

为解决智能船舶远程驾驶控制中心驾驶员负荷精准监测与调控难题，支撑负荷预测模型优化，本数据集以9名远程驾驶员（分为3组，每组3人）为研究对象，采用中科心研 低侵入式腕式可穿戴手环，在简单、中等、困难3种航道场景下采集光电容积脉搏波（PPG）、皮肤电反应（GSR）等原始生理信号。经标准化预处理与特征工程，从原始数据中提取心率变异性、皮肤电活动及脉搏波形态相关的27项核心生理特征；通过XGBoost分类器计算特征重要性，筛选出排名前6的关键特征用于模型构建，并采用网格搜索结合五折交叉验证优化模型超参数，最终模型验证集准确率达0.8696，预测性能稳定。本数据集包含四大核心子集：（1）算法代码子集，实现特征选择、超参数寻优、模型训练与测试功能；（2）模型参数子集，包含特征权重排名 、超参数配置及训练参数；（3）原始数据子集，按航道场景分类存储9名驾驶员的PPG原始信号；（4）特征数据子集，包含生理特征数据、训练数据、测试数据及预测结果 。本数据集实现了远程驾驶任务负荷标签与生理数据的精准关联，可为智能船舶远程驾驶负荷预测模型优化、负荷调控策略验证及人因工程适配性研究提供可靠的数据支撑与算法参考和数据支撑。

To address the challenges of accurate monitoring and regulation of driver workload in the remote driving control center of intelligent ships, and to support the optimization of workload prediction models, this dataset takes 9 remote drivers (divided into 3 groups with 3 people in each group) as research subjects, and uses low-invasive wrist-worn smart wristbands from Zhongke Xinyan to collect original physiological signals such as photoplethysmography (PPG) and galvanic skin response (GSR) under three waterway scenarios: simple, medium and difficult. After standardized preprocessing and feature engineering, 27 core physiological features related to heart rate variability, skin electrical activity and pulse wave morphology are extracted from the original data. The feature importance was calculated using an XGBoost classifier, and the top 6 key features were screened for model construction. Grid search combined with 5-fold cross validation was used to optimize the model hyperparameters, and the final model achieved an accuracy of 0.8696 on the validation set with stable prediction performance. This dataset includes four core subsets: 1. Algorithm code subset: Implements functions such as feature selection, hyperparameter tuning, model training and testing; 2. Model parameter subset: Contains feature weight rankings, hyperparameter configurations and training parameters; 3. Raw data subset: Stores the original PPG signals of the 9 drivers classified by waterway scenarios; 4. Feature data subset: Includes physiological feature data, training data, test data and prediction results. This dataset realizes the accurate association between remote driving task workload labels and physiological data, and can provide reliable data support and algorithmic references for the optimization of intelligent ship remote driving workload prediction models, the verification of workload regulation strategies and the research on human factors engineering adaptability.

搜集汇总

背景与挑战

背景概述

该数据集旨在解决智能船舶远程驾驶中驾驶员负荷监测与调控的难题，通过采集9名驾驶员在三种航道场景下的生理信号（如PPG和GSR），提取27项特征并利用XGBoost算法筛选关键特征，构建了准确率达0.8696的预测模型。数据集包含算法代码、模型参数、原始数据和特征数据四个子集，为智能船舶远程驾驶负荷预测模型优化和人因工程研究提供数据支撑与算法参考。