SmartCorners - Innovative predictive control solutions supported by fleet data (V1) (D5.1)

This report presents the interim results of Work Package 5 (WP5) “Optimal, safe and secure operation on vehicle level” of the SmartCorners project. It emphasizes the development and initial validation of new predictive control solutions for a battery electric vehicle (BEV) equipped with four smart e-corner modules. These modules combine in-wheel motors (IWMs), brake-by-wire, and advanced suspension and kinematic actuators at each wheel. This setup allows for distributed actuation and offers new avenues for energy-efficient, safe, and comfortable vehicle operation throughout its lifespan. In this context, Deliverable D5.1 “Innovative predictive control solutions supported by fleet data” outlines the status at M24 of two main activities: Task T5.1, which develops predictive integrated chassis control for the SmartCorners electric vehicle (EV), and Task T5.2, which enhances vehicle-level control with a mission-management layer that uses fleet data and off-board computation. The findings presented here translate the project's high-level goals of energy efficiency, safety, comfort, robustness, and user value into specific control concepts, quantitative key performance indicators (KPIs), and a control architecture that can be implemented on the demonstrator vehicle and connected test fleets. Section 2 introduces the motivation and technical background, explaining how current EV architectures are limited by separately designed propulsion, braking, suspension, and steering systems. It shows how smart e-corners with IWMs and by-wire interfaces enable more integrated and user-focused control strategies. It also situates WP5 within the entire SmartCorners project, where advanced control and the use of artificial intelligence (AI)-driven data are crucial for fully leveraging the new hardware design. Section 3 summarizes the technical requirements guiding Tasks T5.1 and T5.2, including limitations from the SmartCorners vehicle architecture, multi-criteria control goals (safety, energy efficiency, thermal limits, comfort, tire and component usage, robustness, and real-time feasibility), and how on-board, contextual, and fleet data play a role in supporting predictive and mission-level decisions. Section 4 describes the overall control architecture. It is organized hierarchically into actuator, coordinated chassis/powertrain, supervisory, and mission-management layers, and includes functions for diagnostics, safety, cybersecurity, and data logging. The architecture follows a “local autonomy plus remote intelligence” model: all safety-critical, real-time control loops operate on board, while intensive learning and optimization tasks are carried out off board using cloud resources and fleet data. Clear interfaces are established between the on-board system, the off-board analytics platform, and the digital-twin and X-in-the-Loop (XiL) environments from other WPs, ensuring traceability from simulation to vehicle implementation. Section 5 explains the concepts and intended implementation of predictive integrated chassis control developed in Task T5.1. Based on a full-vehicle model that includes IWMs, active suspension, and wheel-kinematic actuators, the approach uses a nonlinear model predictive controller (NMPC) to coordinate wheel torques, vertical forces, and kinematic angles within a unified optimization framework. Predictive use of map and advanced driver assistance system (ADAS) information enables proactive body control, better torque vectoring (TV), improved grip utilization, and lower energy consumption through rolling-resistance and ride-height optimization. To meet real-time demands on automotive electronic control units (ECUs), the project uses surrogate models based on neural networks (NNs) and imitation learning (IL), allowing NMPC behaviour to be approximated with significantly less computational cost while maintaining accuracy. Section 6 describes the mission-control concepts supported by the fleet and initial implementation steps in Task T5.2. Building on the “Virtual Active Rollbar” concept developed in earlier work, the IWMs generate roll-controlling vertical force vectoring (FV) while a dual-loop architecture combines a fast reactive layer with a predictive roll-control layer. A slip-control function supervises IWM torques, ensuring safe tire-force use and providing continuous estimates of local tire–road friction. These grip estimates, along with comfort-related KPIs and contextual information, are sent to a cloud-based analytics platform and compiled into a “road-intelligence” layer. Vehicles can then get compact road descriptions, such as expected friction and roughness along a route, and use them to adjust predictive roll control ahead of critical segments. This improves ride comfort and consistency while ensuring safety even with unreliable connectivity. Initial tests on the demonstrator confirm that predictive roll control with IWM-based vertical FV is viable on the vehicle and shows comfort improvements in line with previous simulations. Section 7 connects these technical advancements to the overall SmartCorners goals, milestones, and expected impacts. D5.1 is a key verification element for Milestone 4 “Control strategy available” at M24, showing that essential predictive control and fleet-data concepts have moved from design to functioning implementations and initial vehicle validation. Section 8 concludes the report, noting that no critical issues affecting Tasks T5.1 and T5.2 have been identified at this stage and outlines the plan toward Deliverable D5.4 at M34, where the fully integrated, end-to-end concept—including fleet-data collection, cloud-based model updating, and vehicle-side use of road-intelligence inputs—will be demonstrated and reported along with consolidated KPI-based evaluations.

Перевод пока недоступен