This paper addresses the problem of coupon strategy optimization of service providers in ride-sharing platforms. Since passengers prefer service providers offering lower fares, service providers have a strong incentive to use coupon strategies to secure orders. Accordingly, this paper proposes FCA-RL, a novel reinforcement learning-based subsidy strategy framework that rapidly adapts to competitors’ price changes and optimizes order volume under budget constraints. FCA-RL integrates two key techniques: Fast Competition Adaptation (FCA) to accelerate competition adaptation and Reinforced Lagrangian Adjustment (RLA) to optimize coupon decisions while respecting budget constraints. In addition, we introduce RideGym, a dedicated simulation environment for evaluating and benchmarking various pricing strategies. Experimental results show that FCA-RL outperforms existing methods in various market situations.
