Abstract: In high-risk scenarios such as disaster response, robots must operate under severe resource constraints, high environmental uncertainty, and irreversible consequences of action. Their decisions are therefore not governed solely by efficiency and safety, but also by the need to resolve complex ethical value conflicts. To address this challenge, a general ethical decision-making mechanism, which can be seamlessly embedded into any two-stage trajectory planning framework, was proposed. The mechanism operates between coarse trajectory generation and fine trajectory optimization, enabling ethical evaluation and priority ranking of candidate tasks and trajectories without altering the underlying planning algorithm. Specifically, a multi-dimensional ethical value model was first established, and a large language model was leveraged to construct a digital expert-based ethical assessment and value ranking process, through which abstract ethical principles were translated into quantifiable and comparable decision weights with improved consistency and interpretability. Based on this, zero-determinant strategies were introduced to impose formalized ethical constraints on the task selection process, ensuring stable and controllable decision preferences under uncertainty and risk. Simulation results demonstrated that the proposed mechanism significantly enhanced decision transparency and robustness in ethically challenging scenarios while reducing energy consumption and execution risk without compromising rescue success rates. The proposed approach provides a general and reliable solution that balances ethical soundness with executional reliability for autonomous rescue robots.