From fragments of information to evidence. Strengthening collision investigation for software defined vehicles

Modern passenger vehicles are increasingly software-defined, with safety-critical behaviour implemented through networks of electronic control units, sensors, and complex software. When severe incidents occur, such as sudden unintended acceleration (SUA) or unexpected braking, investigators often face fragmentary event logs, incomplete capture of sensor and actuator signals, and limited transparency into manufacturers' safety analyses. We examine these gaps and present an investigation-oriented framework, including a failure-mode taxonomy, to transform "data fragments" into defensible safety evidence. Drawing on real-world cases in which recorded vehicle data indicated high accelerator-pedal (throttle) demand while other evidence pointed to deceleration or limited driver intent, our analysis identifies several plausible electronic and software-related failure modes. These include transient power or sampling anomalies, signal-processing misinterpretation, and unintended system activation. It also shows why short pre-crash windows and sparse sampling hinder causal determination without supporting diagnostics and engineering evidence. Our workflow

• couples engineering reconstruction with Safety Cases and common safety analyses
• motivates an EDR+ oriented set of essential evidence needs for next-generation incident logging and investigation, including pre- and post-event recording duration, sampling rates, and priority signals spanning driver inputs, supervisory functions, power and voltage health, and automated interventions.

Keywords: Collision investigation, Software defined vehicles, Advanced Driver Assistance Systems, Sudden Unintended Acceleration, Event Data Recorder (EDR) and Human-machine interface (HMI).