Resource Allocation in Software-Defined Vehicles: ILP Model Formulation and Solver Evaluation

Modern vehicles are changing rapidly. Years ago, cars were still differentiated from each other by mechanical features such as horsepower. Nowadays, there are nearly 100 electronic control units (ECUs) and millions of lines of codes embedded in a vehicle. In the automotive industry, software is starting to have more value than hardware. Thus, the concept of software-defined vehicles (SDVs) is rising. Traditional electrical and electronic (E/E) architectures show their limitations with predefined implementations, layout and distribution, as well as insufficient computational power for future extensions. Instead, powerful automotive central computing platforms with flexible software solutions are being developed. With the development of hypervisor technology, applications with different safety criticality can be deployed on different virtual machines (VMs) of the same physical hardware. The resource allocation problem, which refers to mapping software to hardware, is becoming more complicated with the increasing number of automotive applications. Existing approaches for resource allocation problems mainly focus on distributed E/E architectures. In this paper, we aim to address the gap of solving resource allocation problems in the central computing platform of SDVs, where the dynamical VM creation and constrained application deployment shall be considered simultaneously. We propose an Integer Linear Programming (ILP) based approach by formulating the problem as an optimization problem with the minimum number of VMs as the goal and utilizing well-known solvers to find the solution automatically. Moreover, we evaluate the performance of state-of-the-art solvers for the proposed approach.