Dataset - Melding LLM and temporal logic for reliable human-swarm collaboration in complex environments

Ablation Experiment Data Data: Ablation Experiment/ Context Analysis.csv: Performance comparison of different LLM models (Qwen variants) with metrics for Success Rate (SR) and Planning Time (PT) across multiple test runs Module Comparison.csv: System ablation study comparing full system ("ours") against configurations with components removed (w/o RAG, w/o MS-p1, w/o MS-p2) and with human assistance (ours w/ Human), measuring Success Rate, Planning Time, Quality, and Task count Subtask Generation.csv: LLM model size ablation comparing performance of different Qwen model sizes (1.7b, 8b) on subtask generation tasks, recording Success Rate, Planning Time, and Quality metrics Baseline Experiment Data Data: Baseline Experiment/ summary.csv: Comparative baseline results from different baseline methods (DCEN_ETE_NG_NV) with Success Rate (SR), Planning Time (PT), Execution Time (ET), Plan Quality (PQ), Instruction Count (IC), and Task Count (TC) ours.csv: Main system results across difficulty levels (EASY/MEDIUM/HARD) with performance metrics, including human-assisted variant Baseline Main Records/: 40+ timestamped experiment run folders containing detailed execution traces and metrics Consistency Comparison/: Multi-run experiments measuring result consistency across repeated trials Robot Num Comparison/: Experiments with different robot team sizes (12 and 24 robots) Temporal Logic Comparison/: Prompt variations (easy/hard/medium) with corresponding task records (records.csv) HCI Experiment Data Data: HCI Experiment/ Informed consent forms/: Research ethics documentation from human subjects records/: Multiple participant study sessions (e.g., haoxin-manual-e-1, zhangshuo-e-1, yuxiao-manual-m-2, etc.) gsr.csv: Galvanic Skin Response (physiological stress/engagement measurements) end_*_state.csv: Robot state trajectories for each end-effector during tasks edge_*_detection.csv: Edge event detection and analysis edge_*_context_analysis.csv: Context understanding performance llm_subtask_generation.json: LLM-generated subtask plans in JSON format llm_subtask_generation_score.csv: Evaluation scores for LLM subtask quality with RAG queries and scores Simulation Experiment Data Data: Simulation Experiment/ end_*_state.csv (100+ files): Robot state trajectories recording timestamps, robot states (IDLE/MOVE/WAIT), and subtask execution details llm_subtask_generation.jsonl: LLM subtask generation log in JSONL format Hardware Experiment Data Data: Hardware Experiment/ log/: System runtime logs from hardware deployment (10+ drone units) odom/: ROS bag files containing odometry data (motion and localization measurements) for each drone (drone_1_odom.bag, drone_2_odom.bag, etc.) Processed Data Data: Processed data/ efficiency/ time.csv: Planning time metrics by method and difficulty level (mean and std values) memory.csv: Memory usage statistics (Average Task Count and Std) tasks.csv: Task completion statistics human_in_loop.csv: Performance with human assistance metrics explain/ quality.csv: Model explanation quality scores by method and difficulty length.csv: Explanation length analysis predict.csv: Prediction metrics hci/ ablation.csv: Ablation study results from human studies (Expert group performance across different modules with efficiency metrics) train.csv: Training data from HCI experiments radar.csv: Radar chart data for multi-dimensional performance comparison pressure.csv: Workload pressure measurements from human subjects