基于沖突避讓的多智能體有效旁路規(guī)劃

打開文本圖片集

關鍵詞：基于沖突搜索；狹窄路段；有效旁路；多智能體路徑規(guī)劃；臨時等待位置；次優(yōu)路徑中圖分類號：TP242 文獻標志碼：A 文章編號：1001-3695（2025）07-024-2103-06doi：10.19734/j.issn.1001-3695.2024.11.0469

Abstract：Tosolvetheproblemthatmultipleagents innarrowenvironmentcannotavoidtheconflictonkeychannels，which leads to thelow eficiencyof multi-agent path finding，this paper proposed an effective bypassplanning algorithm basedon fieldofviewconflictsearch.Intheupperlayerof theconflict-basedsearchalgorithmframework，thefieldofviewrepresented thecommunicationrangeof theagent.Byscanning the multi-value decision diagramof allagents inthecommunicationrange of theagent，thealgorithm determinedwhether therewasacompetion forsharedresources betweentheagents，leading to conflicts.Themulti-valueddecisionbacktrackingconflictavoidance strategyresolvedtheconflictsbetween multipleagentson thecriticalchannel.Byscanning the multi-valueddecision diagram，theagentobstructingthecritical channelcould find the suboptimalpathorthetemporarywaiting positionasaneffetivebypasstoavoidthecollsion.Theresultsofdifferentscaleand structure maptestshow thatmuli-valued decision backtracking conflictavoidance strategycan effectivelyreducetheoccurrenceof conflictbetwenagents.Intherandom-64-64-10and maze-128-128-10mapenviroments，theaveragesuccessrateof pathfinding for multiple agents improves by 10% and 15.2% ，respectively，compared with CBS-BP，proving that this algorithm effectively avoids conflicts and increases pathfinding success rates.

Key words：conflict-based search；narrowsection；efectivebypass；multi-agentpath planning；temporary waiting position; suboptimal path

0 引言

多智能體路徑規(guī)劃（multi-agentpath finding，MAPF）作為智能體領域的研究熱點，受到國內(nèi)外學者廣泛關注[1-3]。（剩余14188字）