Dependency-aware OTA execution planning with path reuse for software-defined vehicles

초록

As vehicle platforms become more centralized and software-defined, over-the-air (OTA) updates increasingly act on coupled vehicle-level software configurations rather than isolated electronic control units (ECUs). ECU-version transitions and cross-ECU compatibility constraints make OTA execution a problem of selecting feasible intermediate vehicle states. We formulate this problem as dependency-aware vehicle-state graph search, where each state is a complete multi-ECU software configuration and each action updates one ECU version while preserving dependency feasibility. The formulation accounts for release-aligned targets and transmitted-volume costs for full-image and differential update artifacts. Rather than introducing a new generic search algorithm, we evaluate the formulation using practice-oriented execution policies, exact reference search, and standard A* search. We also study execution-path reuse for repeated OTA campaigns with partial deployments, maintenance, component replacement, or evolving target releases. Controlled simulations show two complementary effects. Standard A* search reduced transmitted data volume by 8.2%-66.9% relative to execution-policy baselines while revealing multi-step and repeated-ECU update behavior. In related-instance scenarios, execution-path reuse reduced total action evaluations by more than 98% relative to from-scratch replanning, but can trade computational efficiency against transmitted data volume. These results support treating SDV OTA as dependency-aware in-vehicle execution planning with reusable execution paths.

키워드