Multi-source network-level pavement friction measurements via instrumented and dynamically routed autonomous vehicles

• Dynamic compilation of network pavement friction and slip ratio measurements from instrumented regular and autonomous vehicles. • Systematic correction of vehicle-specific bias and reduction of variances from fused data. • Safety impacts assessment based on pavement friction measurement accuracy. • Dynamic AV routing based on real-time updates of network friction measurements in a rolling horizon framework. • Routing problem solved by customized column generation and branch-and-price algorithms. This paper proposes a new framework for network pavement friction measurements based on dynamically compiling acceleration and slip ratio data from instrumented vehicles. These vehicles include regular vehicles (RVs) driven by the public, and a fleet of autonomous vehicles (AVs) that are centrally controlled by the transportation agency to perform a broader range of acceleration and deceleration activities. Statistical analysis is conducted to reveal (i) how measurements from different vehicles shall be fused into parametric friction models to ensure precision and efficiency of pavement friction estimates, and (ii) how real-time updates of network friction measurements can guide AV routing in a rolling horizon framework to enhance measurements at the most needed spots. This proposed modeling framework is tested in a prototype case study based on a simulated highway network in Champaign County, Illinois. Initial results showed that relying solely on uncontrolled RVs provided sub-linear convergence on pavement friction measurement precision over time. In contrast, using even a few controlled AVs could quickly cover hundreds of miles of roadway segments within a few hours, and significantly improve friction estimates across the network.