Abstract: As aerospace threats continue to evolve toward diversification, intelligentization, and systematization at an accelerating pace, the limitations of traditional radar networks in heterogeneous semantic interoperability, dynamic environment adaptation, and extreme-condition survivability have become increasingly prominent. To address these challenges, this paper proposes an autonomous collaborative architecture for radar networks in complex battlefield environments, based on a four-layer progressive framework of Elastic Connectivity, Unified Semantics, Decision-making and Control, and Service Provision. By integrating elastic networking, dynamic ontology modeling, and hierarchical decision-making and control, the architecture establishes a closed-loop operational mechanism of Connectivity, Understanding, Decision-making, and Service, systematically resolves the core bottlenecks of heterogeneous interoperability, insufficient generalization in dynamic scenarios, and structural vulnerability under intense countermeasures, and drives radar networks toward a paradigm shift from discrete nodes to collective intelligence. This work provides a technical path that balances theoretical innovation with engineering feasibility for the construction of a new-generation intelligent, resilient, and service-oriented aerospace early warning system.