An FPGA-Oriented Offline–Online Framework for Multi-Robot Task Allocation

With the increasing deployment of multi-robot systems in autonomous applications, efficient task allocation is essential for coordinating multiple robots and improving system performance. To address this challenge, this paper proposes an FPGA-oriented offline–online MRTA framework that transforms online combinatorial optimization into offline decision-space compilation and hardware-efficient online strategy selection. An interval-robust dominance criterion is developed to prune strategies that cannot become optimal within a prescribed energy-cost range defined during the offline stage. Pairwise comparisons among the retained strategies are then compiled into linear decision boundaries, reducing online strategy selection to linear classification. The resulting mechanism is implemented on a Xilinx Zynq-7010 FPGA using a lightweight multiplier-free accumulation architecture. Experimental results demonstrate reduced decision latency and deterministic embedded execution. The main contribution of this study is a hardware-oriented MRTA method that replaces online combinatorial search with bounded arithmetic, comparison, and control operations suitable for FPGA implementation.