We introduce a new approach to address the task allocation problem in a system of heterogeneous robots comprising of Unmanned Ground Vehicles (UGVs) and Unmanned Aerial Vehicles (UAVs). The proposed model, \texttt{\method}, or \textbf{G}raph \textbf{A}ttention \textbf{T}ask \textbf{A}llocato\textbf{R} aggregates information from neighbors in the multi-robot system, with the aim of achieving joint optimality in the target localization efficiency.Being decentralized, our method is highly robust and adaptable to situations where collaborators may change over time, ensuring the continuity of the mission. We also proposed heterogeneity-aware preprocessing to let all the different types of robots collaborate with a uniform model.The experimental results demonstrate the effectiveness and scalability of the proposed approach in a range of simulated scenarios. The model can allocate targets' positions close to the expert algorithm's result, with a median spatial gap less than a unit length. This approach can be used in multi-robot systems deployed in search and rescue missions, environmental monitoring, and disaster response.

翻译：本文提出了一种解决异构机器人系统（包括无人地面车辆和无人航空器）中任务分配问题的新方法。所提出的模型名为GATAR（图注意力任务分配器），通过聚合多机器人系统中相邻节点的信息，旨在实现目标定位效率的联合最优。该方法采用去中心化架构，具有高度鲁棒性和自适应性，能够应对协作对象随时间动态变化的情况，确保任务连续性。我们还提出了异构感知预处理技术，使所有不同类型的机器人能够通过统一模型进行协作。实验结果表明，该方法在一系列仿真场景中具有有效性和可扩展性。该模型分配的目标位置接近专家算法的结果，中位空间差距小于一个单位长度。该方法可应用于搜索救援、环境监测和灾难响应等任务中的多机器人系统。