Affinity-Aware Offloading is a kernel offloading method that selectively assigns tasks to processing units based on computational affinity, prioritizing resource locality and data proximity to minimize overhead and optimize performance.