<p>Coverage path planning (CPP) directs a robot to traverse all accessible areas of a domain while avoiding obstacles. This research considers a mobile robot operating on uneven 3D terrain, where elevation changes make path generation more complex than on flat surfaces. A real-world CPP must balance multiple objectives; for example, in agriculture, a tractor should cover the entire field while minimizing fuel consumption. Since no single path optimally satisfies all goals, CPP is naturally formulated as a multi-objective optimization (MOO) problem. The existing methods perform inadequately under these conditions. The terrain adaptive NSGA-II (TA-NSGA-II) is a modified variant of the NSGA-II that was developed to address this challenge. It optimizes three objectives, namely energy consumption due to (1) total travel distance, (2) cumulative elevation changes, and (3) the number of turns, simultaneously and independently, while ensuring 100% coverage of the target area. The approach is evaluated on synthetic terrain modeled by Himmelblau’s function and validated on real-world 3D terrain data from Nelson City, New Zealand, and two other uneven sites. TA-NSGA-II is compared with five relevant existing methods: terrain-shape adaptive (TSA), zigzag, Gaussian bioinspired neural network (GBNN), spanning tree coverage (STC), and the modified NSGA-II. The TA-NSGA-II method outperforms existing approaches, achieving a 5–237% reduction in energy consumption while maintaining 100% coverage. An ablation study underscores the importance of elevation objectives and constraints. The generated Pareto optimal paths enable selection on the basis of trade-offs among energy consumption, number of turns, elevation change, and travel distance.</p>