High-precision trajectory tracking for wheeled mobile robots is hampered by parameter uncertainties and external disturbances. While Sliding Mode Control (SMC) offers inherent robustness, its performance is limited by fixed parameters that cannot adapt to varying trajectory dynamics, creating a trade-off between accuracy and stability. This paper proposes a Reinforcement Learning-enhanced Adaptive Sliding Mode Control (RL-SMC) framework to overcome this limitation. The architecture employs a reinforcement learning agent for the online optimization of crucial SMC parameters, such as control gains and sliding surface coefficients. A key innovation is the integration of a trajectory preview mechanism, which provides the agent with foresight into upcoming path curvature, enabling proactive and context-aware adaptation. Extensive simulations on dynamically challenging trajectories reveal that the proposed controller achieves superior tracking accuracy and robustness compared to both fixed-parameter SMC and pure RL approaches. The framework also demonstrates enhanced stability against system noise and provides significantly smoother control actuation, validating its ability to achieve both high precision and adaptive stability.