Empirical performance criteria based on geometric and physical properties are commonly used for real-time decision-making (and design) of redundant anthropomorphic manipulators (serial) or workcells (parallel) performing high level tasks while avoiding obstacles, providing safety, and responding to human commands. This requires configuration management of these complex systems by prioritizing candidate configurations relative to quantifiable secondary objectives with clear physical meanings. Clarity is addressed by using position, inertial, kinetic, and potential energy (gravity and deformation) based metrics that are crisply defined from system and input parameters. Scaling differences among derived performance metrics require normalization to determine their relative import and inclusion in multicriteria decision-making techniques. Large dimensional spaces mean statistically reduced representations are necessary to decipher their relative import and allow an operator without extensive robotic knowledge to use them effectively and independent of a redundancy resolution technique (RRT). We propose two global norms: the unit norm and the average norm for a broad set of performance criteria that quantify these systems’ unique constraint, transmission, and energy characteristics. These norms are then used to select relevant criteria for operational decision-making based on intuitive operator commands instead of abstract mathematical notions. Additionally, a modified RRT is presented that is more robust with respect to changes during operation in the secondary objectives, which allows for greater flexibility when formulating new criteria. Results are illustrated using a variety of new and existing criteria on an anthropomorphic dual-arm system.