Autonomous relative navigation is an enabling capability for the next generation of distributed space missions, including satellite swarms for target inspection, sparse synthetic apertures, on-orbit assembly, and interferometric sensing. These mission concepts require infrastructure-free, onboard relative state knowledge to coordinate maneuvers and maintain geometry. Yet many operational concepts still rely on GNSS availability, centralized ground processing, or a designated chief spacecraft that aggregates measurements. These approaches limit autonomy, reduce scalability, and can become a single point of failure. Inter-satellite radio-frequency (RF) ranging is an attractive alternative because it enables joint communication and metrology without external infrastructure and without imposing strict constraints on the platform and mission design. To date, distributed estimation methods that explicitly account for relative state observability, link topology, and clocks misalignment effects remain underdeveloped. 

This paper presents an estimation filter and ranging signal co-design framework for distributed 3-DoF relative navigation in spacecraft formations using inter-satellite RF range and range-rate measurements. The framework, developed by the ASTRA research laboratory at Politecnico di Milano, targets decentralized operation: each spacecraft maintains a local estimate of the relative state and refines it through limited message passing with neighboring agents. The approach couples observability-aware distributed filtering and joint communication–ranging and time-synchronization methods, enabling scalable relative navigation in multi-agent swarms. 

On the estimation side, relative navigation is formulated as a networked stochastic filtering problem driven by nonlinear range and range-rate observations. Two different families of distributed filters are developed. First, Kalman consensus architectures combine local innovations with consensus terms over dynamic communication graphs, allowing each agent to benefit from neighborhood information while keeping computational complexity bounded. Second, information-weighted consensus filters are developed to support fusion under limited bandwidth and to reduce sensitivity to correlated information. For both families, stability and convergence are analyzed with respect to graph connectivity and measurement availability; state-sharing versus measurement-sharing strategies that guarantee convergence under realistic link schedules are also investigated. Central to the framework is an observability characterization that connects formation geometry, relative motion, and communication topology to the rank and conditioning of the distributed estimation problem, providing criteria for selecting neighbors and scheduling updates. 

On the measurement and networking side, RF ranging and communication are treated as inseparable from multi-access constraints and onboard clock behavior. We propose integrated signaling options (TDMA, CDMA, and hybrid schedules) in which a single waveform supports both data exchange and ranging, enabling simultaneous coordination and navigation while controlling interference and spectrum usage. The effects of clock bias and drift are analyzed, and synchronization strategies are introduced to mitigate timing errors that otherwise map directly into range and range-rate biases. 

End-to-end performance is assessed through simulation campaigns across representative formation geometries and time-varying communication graphs motivated by line-of-sight constraints and link scheduling. Metrics include relative position and velocity reconstruction accuracy, convergence time, sensitivity to initialization, robustness to intermittent measurements and packet loss, and realistic RF impairments (SNR variations, multipath, and non-ideal Doppler extraction). The results highlight that embedding observability constraints into the distributed filter design prevents divergence in weakly observable configurations, and that timing-aware processing substantially improves accuracy when low-cost oscillators are assumed. The transfer of these methods to embedded SDR/SoC implementation and multi-agent experimental validation is currently under development.