Chemical looping air separation (CLAS) represents a promising approach for efficient O2 production from the air. The present study aims at optimizing the absorber/desorber operations and the separation process with extensive experimental validation. Specifically, a one-dimensional packed bed model was developed to investigate the CLAS operation with a Sr0.8Ca0.2Fe0.9Co0.1O3-δ perovskite sorbent. The sorbent's redox thermodynamics was measured by thermogravimetric analysis. The results were subsequently incorporated into a linear driving force model to describe the O2 absorption and desorption rates. Both 4-step and 5-step cycle configurations, with various cyclic structures, were performed in a subpilot-scale packed bed. The model predicted O2 purity and productivity were consistent with experimental results, supporting its accuracy and applicability. Parametric analysis and multi-objective optimization were further conducted to assess the performance of CLAS. Both O2 purity and recovery increased monotonically with the cycle time, airflow rate, steam flow rate, and absorption pressure. Meanwhile, optimal O2 productivity and power consumption can only be achieved by specific combinations of these parameters. The optimized results showed that CLAS can be highly competitive when compared to conventional pressure swing adsorption (PSA) or cryogenic distillation. The 5-step cycle configuration achieved a minimum power consumption of 118 kW·h for producing 1 ton O2 with ≥95% purity. The maximum O2 productivity reached 0.0932 gO2/(gsorbent·h) with 390 kW·h/ton O2 of energy consumption (95% pure). The optimization results also indicate that CLAS can potentially be more efficient than cryogenic distillation even when the required O2 purity is above 99%.