Autonomous remotely operated vehicles (ROVs) have emerged as a promising solution for fish pen inspections, replacing labor-intensive manual inspections and enhancing operational efficiency. However, deploying ROVs in these environments presents significant challenges, including unreliable localization, system uncertainties, and dynamic environmental disturbances. To address these issues, we propose a fully autonomous ROV system designed specifically for fish pen inspection in this paper. The system integrates a vision-aided approach for underwater localization and an effective path-planning algorithm to ensure safe navigation during inspections. To improve the system’s robustness against uncertainties and disturbances, we introduce a robust control scheme that combines two components: a nominal feedback controller that stabilizes the partially known nominal model of the ROV dynamics, and a sliding-mode compensator (SMC) that mitigates the effects of unknown dynamics and external disturbances. This robust control scheme, referred to as RC-SMC, minimizes the need for extensive parameter tuning while ensuring precise path-tracking underwater. Comprehensive experiments have been conducted in both laboratory and field environments to validate the efficacy and robustness of the proposed system. The results demonstrate that our ROV system can effectively perform autonomous inspections while maintaining improved stability and tracking precision compared to existing algorithms in the presence of various uncertainties and disturbances.