摘要：Task allocation under uncertain conditions is a keyproblem for agents attempting to achieve harmony in disaster environments.This paper presents an agent-based simulation to investigate task allocationconsidering appropriate spatial strategies to manage uncertainty in urbansearch and rescue (USAR) operations. The proposed method is based on thecontract net protocol (CNP) and implemented over five phases: orderingexisting tasks considering intrinsic interval uncertainty, finding acoordinating agent, holding an auction, applying allocation strategies (fourstrategies), and implementing and observing the real environment. Applyingallocation strategies is the main innovation of the method. The methodologywas evaluated in Tehran's District 1 for 6.6, 6.9, and 7.2 magnitudeearthquakes. The simulation began by calculating the numbers of injuredindividuals, which were 28 856, 73 195, and 111 463 people for eachearthquake, respectively. Simulations were performed for each scenario for avariety of rescuers (1000, 1500, and 2000 rescuers). In comparison with theCNP, the standard duration of rescue operations with the proposed approachexhibited at least 13 % improvement, with a maximal improvement of 21 %.Interval uncertainty analysis and comparison of the proposed strategiesshowed that increased uncertainty led to increased rescue time for the CNPand strategies 1 to 4. The time increase was less with the uniformdistribution strategy (strategy 4) than with the other strategies. Theconsideration of strategies in the task allocation process, especiallyspatial strategies, facilitated both optimization and increased flexibilityof the allocation. It also improved conditions for fault tolerance andagent-based cooperation stability in the USAR simulation system.
