Real-time traffic signal control (TSC) in road networks remains challenging due to variable traffic flows and high computational complexity. Existing model predictive control (MPC) approaches often face several limitations, including the reliance on commercial solvers, the inflexibility of single-objective optimization, and the use of prediction models that fail to leverage rich data from existing sensors. In light of these limitations, this paper aims to introduce a license plate recognition (LPR) data-driven, distributed MPC approach for efficient real-time network TSC with multi-objective optimization. The approach implements a multi-agent architecture where each intersection operates as an independent agent while maintaining network-wide coordination through minimal information exchange with neighboring intersections. For vehicle arrival prediction, a platoon dispersion model is combined with a feedback mechanism that leverages LPR data. This integration enhances prediction accuracy and mitigates error accumulation in the rolling horizon scheme. For signal timing optimization, a multi-objective dynamic programming algorithm that incorporates Pareto optimality was developed to dynamically balance vehicle delay and queue length, providing greater adaptability to changing traffic conditions. Evaluations in a (Formula presented.) grid network demonstrate that the approach reduced delays by up to 11%, compared to Synchro and 46% compared to max pressure control. The scalability and practical applicability were further validated through implementation in a real-world network of 28 intersections. Crucially, the approach is computationally efficient, with median optimization times remaining under 0.05 s per intersection in the real-world network, confirming its suitability for large-scale, real-time implementation.