Scholarlink

Novel Earth observing systems aim to leverage newly developed Uncrewed Aerial Vehicles (UAVs) to provide higher resolution spatio-temporal data than is currently available through satellite, crewed flights, or in-situ sensors, for many climate science applications. The Intelligent Long Endurance Earth Observing System (ILEOS) provides science activity planning for obtaining higher quality spatio-temporal data from UAVs. ILEOS fuses coarse-grained data available from satellites and other sources to identify high value science targets, and produces flight plans for high altitude long endurance (HALE) UAVs, choosing which targets to observe to maximize science return. ILEOS uses a variant of the Orienteering Problem with time-dependent science reward values, with novel extensions to model vehicle turn times and no fly zones. Two solution methods are presented: Mixed Integer Linear Programming (MILP) and Monte Carlo Tree Search (MCTS). Experiments are conducted to compare their performance on six real-world scenarios. Results show that MCTS often outperforms a leading MILP solver.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Time-Dependent Orienteering for High Altitude UAVs to Monitor Greenhouse Gases:

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Richard Levinson,
  • Vinay Ravindra,
  • Jeremy Frank,
  • Meghan Saephan

摘要

Novel Earth observing systems aim to leverage newly developed Uncrewed Aerial Vehicles (UAVs) to provide higher resolution spatio-temporal data than is currently available through satellite, crewed flights, or in-situ sensors, for many climate science applications. The Intelligent Long Endurance Earth Observing System (ILEOS) provides science activity planning for obtaining higher quality spatio-temporal data from UAVs. ILEOS fuses coarse-grained data available from satellites and other sources to identify high value science targets, and produces flight plans for high altitude long endurance (HALE) UAVs, choosing which targets to observe to maximize science return. ILEOS uses a variant of the Orienteering Problem with time-dependent science reward values, with novel extensions to model vehicle turn times and no fly zones. Two solution methods are presented: Mixed Integer Linear Programming (MILP) and Monte Carlo Tree Search (MCTS). Experiments are conducted to compare their performance on six real-world scenarios. Results show that MCTS often outperforms a leading MILP solver.