Full results for "Swarm Autonomy Simulation Thesis" (all episodes)
收藏资源简介:
Complete episode-level outputs for all simulation sweeps used in the thesis "Autonomous Drone Decision-Making: Micro-Decisions, Meaningful Human Control, and Governance toward 2035". This dataset contains:- All 24,330+ episode JSON files, compressed into streaming-friendly .jsonl.gz blocks (organized by sweep)- Aggregated CSV files (sim_aggregates_FINAL.csv with 24k episodes × 45 metrics/config columns)- Log files per sweep (compressed)- Sample JSON files for quick inspection- Complete manifest and checksums for integrity verification Episodes are provided as compressed JSONL blocks (~10,000 episodes per block) for efficient download and streaming. Each sweep (A_MHC_core_max, B_HITL_max, C_ROE_operational, C_ROE_named, D_hostiles_max, E_world_max) is organized in separate directories. For quick analyses, use the aggregated CSV files. For episode-level analyses, extract and process the .jsonl.gz blocks. Total size: ~1.0 GB (compressed).
本数据集包含论文《自主无人机决策:面向2035年的微观决策、有效人为管控与治理》中所用全部仿真参数扫描(simulation sweep)的完整回合(episode)级输出结果。 本数据集涵盖以下内容: 1. 共计24330余份回合级JSON文件,已压缩为适配流式读取的.jsonl.gz分块文件(按参数扫描任务分组); 2. 聚合型CSV文件(sim_aggregates_FINAL.csv,包含24000条回合数据,共45项指标/配置列); 3. 各参数扫描任务对应的压缩日志文件; 4. 用于快速预览的示例JSON文件; 5. 用于完整性校验的完整清单与校验和文件。 所有回合数据以压缩JSONL分块形式存储(每块约含10000条回合数据),以优化下载与流式读取效率。所有参数扫描任务(A_MHC_core_max、B_HITL_max、C_ROE_operational、C_ROE_named、D_hostiles_max、E_world_max)均分别存放于独立目录中。 若需开展快速分析,可直接使用聚合型CSV文件;若需开展回合级精细化分析,则需提取并处理.jsonl.gz分块文件。 总压缩体积约为1.0 GB。



