Human robot collaboration is moving from pilot cells to routine production as manufacturers seek flexible automation that can coexist with skilled work, frequent changeovers, and high product variety. Industry 5.0 further emphasizes human centric system performance, where worker well-being and long-term health outcomes are treated as design requirements. In many production settings, ergonomic risk is dynamic. It changes with layout, pace, product mix, and within shift fatigue. As a result, one-time ergonomic assessments are often insufficient for collaborative workcells where robot timing and handover geometry can influence sustained reaches and repeated trunk flexion. Ergonomics sensing offers a practical pathway by estimating posture, exposure, and related workload indicators in real time and enabling collaboration policies that adapt to changing human state. This paper synthesizes sensing technologies used in collaborative environments, including wearable, vision and depth, robot embedded, and multimodal configurations. It then organizes ergonomic metrics and estimation pipelines that translate raw signals into interpretable outputs such as posture risk indices, exposure measures, fatigue proxies, and workload indicators. The paper also analyzes how sensing can improve collaboration through robot motion adaptation, task allocation, and workcell commissioning decisions, while considering safety constraints and deployment barriers such as occlusion, privacy, and usability. Finally, a practical framework is proposed to connect modular sensing inputs to ergonomic state estimation, decision making, and verification, with the goal of supporting more consistent reporting and more actionable translation from measurement to intervention in Industry 5.0 human-robot collaboration (HRC).