Adaptive Mission Optimization for Electric Vehicles Data-Driven Algorithms for Real-Time Monitoring, Driver Eco-Coaching and Cloud-Edge Integration

The rapid shift toward Battery Electric Vehicles (BEVs) makes every consumed kilowatt-hour critical: on the same route, a sporty driving style can drain the State of Charge (SoC) far faster than an eco approach, with direct repercussions on cost and emissions. This thesis demonstrates that a coordinated set of algorithms for mission optimization, real-time monitoring, and adaptive coaching can steer drivers toward more efficient behaviors, substantially cutting energy demand without sacrificing comfort or increasing travel time. The mission-optimization algorithm, developed in MATLAB/Simulink, computes once per mission an energy-optimal speed profile. An adaptability layer automatically re-tunes the cost weights according to comfort metrics extracted from the driver’s history, keeping the profile consistent with individual habits. Two lightweight, synergistic modules have been implemented for on-board execution under hard real-time constraints. The Mission Monitoring Algorithm ingests driver commands, longitudinal dynamics, and power-train/battery signals, compares them with the optimal profile received from the cloud, and derives KPIs that express—in real time—the efficiency of the driver’s mission. These indicators feed cloud-hosted post-mission dashboards, giving drivers and fleet managers an instant overview of consumption and savings. Using the same inputs, the Suggestion Algorithm converts the gap between actual behavior and optimal profile into concise, adaptive guidance for the driver, steadily steering energy demand downward without distorting the personal driving style. Human–machine interaction was assessed through a Driver-in-the-Loop campaign on a simulator (SCANeR): four testers with heterogeneous styles cut average energy draw from 5.01% to 3.79% SoC—more than 5% absolute (−24% relative) savings—while trip time increased negligibly; in the most demanding scenarios the improvement exceeded 6.5%. In parallel, a Hardware-in-the-Loop validation on a dSPACE SCALEXIO platform configured to emulate a production ECU with realistic CAN traffic showed that the monitoring and coaching modules run within the hard real-time requirements typical of automotive control units, indicating sound prerequisites for future in-vehicle integration. Overall, the end-to-end architecture shows that intelligent interaction between offline optimization, real-time monitoring, and adaptive coaching can deliver measurable energy savings without compromising the driving experience.