Sensor-assisted cooperative localization and communication in multi-agent networks

This thesis presents research advances on cooperative localization and communication. These two macro trends are investigated in the general context of mobile multi-agent networks for situational awareness applications, where time-varying agents of unknown absolute location are asked to fulfill positioning and information sharing tasks.
The research on localization aims to develop an integrated solution where cooperative self-localization of agents is combined with multitarget detection, localization and tracking. Targets are not only considered as unknown objects to be just localized, but their statistical properties represent valuable information to be used upside-down to refine the agents’ positioning coming from self-definition of the multi-agent network. The operating framework is based on belief propagation, a massage passing algorithm working on the factor graph describing all relationships
among agents, measurements and external variables. The final result is the development of a generalized solution that is flexible enough to accommodate heterogeneous measurements of diverse agent types, ensuring a wide range of applications. The two primary scenarios addressed in this thesis are maritime and vehicular uses cases, but internet of things in industrial and surveillance contexts are other major applications of interest.
The research on communication is focused on vehicular applications, where the use of Vehicle-to-Everything (V2X) links to extend vehicle’s sensing capability by establishing connections with nearby users represents an upgrade for a safer mobility. Such paradigm is here reverted: this thesis proposes the use of vehicles’ sensors to enhance communication performances. The specific research topic consists in developing sensor-assisted beam alignment techniques, where narrow
directional V2X beams are steered according to the information extracted from perception sensors. Attention is given not only to the specific technical aspects of the methodology and technologies, namely Millimeter-Wave (mmWave) and Free-Space Optics (FSO), but a system-level architecture is proposed too. Inter-vehicle cooperation and intra-vehicle sensor data fusion are combined in a unified system targeted to guarantee reliable beam-based V2X links in mobility scenarios.
Both macro-researches deal with a tight integration of heterogeneous sensors with communication in dynamic multi-agent systems. The volatility of interconnections among agents due to mobility and instability of links mandatorily calls for flexible and adaptive techniques, capable of profitably fuse diverse types of information. The outcomes of this thesis demonstrate how a statistical approach is capable of handling realistic problems and developing versatile solutions.