日本フィジカルAI新聞

世界のフィジカルAIを、日本語で。

群制御arXiv:1911.09137

不確実環境下での異種マルチエージェント自律探索のための運動制御

Motion control for autonomous heterogeneous multi-agent area search in uncertain conditions

シェア:XThreadsFacebookLINEはてブBluesky

熱方程式駆動型の確率的探索手法HEDACを用い、運動・センシング特性が異なる複数ロボットの探索運動を集中制御する手法を提案し、従来法より約2倍効率的な探索を実現した。

著者: Stefan Ivić

分類: math.OC, cs.MA, cs.RO

原文アブストラクト

Using multiple mobile robots in search missions offers a lot of benefits, but one needs a suitable and competent motion control algorithm which is able to consider sensors characteristics, the uncertainty of target detection and complexity of needed maneuvers in order to make a multi-agent search autonomous. This paper provides a methodology for an autonomous two-dimensional search using multiple unmanned search agents. The proposed methodology relies on an accurate calculation of target occurrence probability distribution based on the initial estimated target distribution and continuous action of spatial variant search agent sensors. The core of the autonomous search process is a high-level motion control for multiple search agents which utilizes the probabilistic model of target occurrence via Heat Equation Driven Area Coverage (HEDAC) method. This centralized motion control algorithm is tailored for handling a group of search agents which are heterogeneous in both motion and sensing characteristics. The motion of agents is directed by the gradient of the potential field which provides near-ergodic exploration of the search space. The proposed method is tested on three realistic search mission simulations and compared with three alternative methods, where HEDAC outperforms all alternatives in all tests. Conventional search strategies need about double the time to achieve proportionate detection rate when compared to HEDAC controlled search. The scalability test showed that increasing the number of HEDAC controlled search agents, although somewhat deteriorating the search efficiency, provides needed speed-up of the search. This study shows the flexibility and competence of the proposed method and gives a strong foundation for possible real-world applications.

関連論文

  1. 1Boston Dynamics、車輪付きAtlasの研究試作機を紹介研究・技術
  2. 2newmoとKDDIが資本業務提携 2028年に堺市で自動運転タクシー商用化目指す提携・買収
  3. 3Google DeepMind、ロボット向けAIを発表新製品・発表
  4. 4NVIDIA、CoRLでGR00T 2とIsaac Lab 3.0を発表新製品・発表
  5. 5GXOとApptronik、Apolloの物流現場試験を拡大 最大1,000台の展開を構想量産・展開

PR・アフィリエイト広告

FRONTIER(GPU搭載BTOパソコン)

FRONTIER(GPU搭載BTOパソコン)の公式広告バナーFRONTIERでGPU搭載PCを見る ↗
PR本紙発行元 EmplifAI