This article introduces an innovative method for designing a controller for vehicle suspension systems featuring magnetorheological (MR) dampers. The approach revolves around the idea that an optimal control strategy can be achieved by applying a reinforcement learning algorithm with continuous states and actions, utilizing data from real-world or simulated experiments. The Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm is used to process sensor data and determine the appropriate actuation voltage for the MR damper. To assess the system’s effectiveness, a quarter vehicle model is employed, incorporating the modified Bouc-Wen MR damper model. This model enables the calculation of key suspension metrics, including displacement, sprung mass acceleration, and dynamic tire load within the suspension workspace. The reward function in the deep reinforcement learning algorithm is based on the sprung mass acceleration. Results from numerous simulated experiments demonstrate that this approach outperforms traditional suspension control methods in terms of both ride comfort and stability.