<p>In this paper, we investigate a reconfigurable intelligent surface (RIS)-assisted multiuser multiple input single output (MU-MISO) symbiotic radio system that incorporates hardware impairments in the RIS. This paper is aimed at solving the optimization problem to maximize the primary transmission rate while guaranteeing the rate of the secondary transmission in the RIS-assisted MU-MISO system. To this end, we formulate an optimization model that considers the transmit power and RIS phase shift constraints as well as the rate constraint of the secondary transmission. This joint optimization problem is complex and coupled, which is difficult to solve. To tackle this issue, we transform the original optimization problem into a Markov decision process characterized by a mixed-signal reward function. To enhance the reward outcome, we propose a novel algorithm based on soft actor-critic (SAC) reset. Simulation results demonstrate that that the proposed SAC-reset method can achieve a higher average reward compared with the conventional SAC schemes and other state-of-the-art deep reinforcement learning (DRL) algorithms.</p>