Chaos, a ubiquitous phenomenon in nature, has long captivated researchers’ attention regarding its control. Previous studies have indeed proposed a plethora of effective control strategies; however, these conventional approaches invariably require the user to possess a substantial reservoir of prior knowledge about chaotic systems. In this investigation, we introduce a novel control methodology for chaotic systems, leveraging the prowess of deep reinforcement learning (DRL), specifically employing the proximal policy optimization (PPO) algorithm to govern the Hindmarsh-Rose (H-R) model. Our empirical results reveal that the controller, meticulously honed through our training regimen, exhibits remarkable proficiency in steering the H-R model toward a predetermined stable equilibrium. Moreover, we have taken into account the impact of stochastic noise on control performance and conducted comparative experiments. The outcomes unequivocally demonstrate that controllers trained amidst the presence of noise exhibit superior control efficacy and heightened environmental adaptability.