Open, Reproducible and Trustworthy Robot-Based Experiments with Virtual Labs and Digital-Twin-Based Execution Tracing

Eventually, the audit trail is generated as a concise documentation of the robot activities. We refer to [26] for a detailed overview and evaluation of the TraceBot framework on a sterility testing usecase [27]. D. Discussion This semantics-and simulation-driven approach to execution tracing provides several methodological advantages for reproducible robot science. The framework generates complete documentation of robot reasoning processes, eliminating the black-box problem that many automated systems suffer from. Beyond documentation, semantic and cognitive traces allow researchers to understand not just what the robot did, but why it made specific decisions and how it arrived at execution outcomes. In future work, we are investigating imagination-enabled traces for automated formal verification of task execution, enabling detection of protocol deviations or unexpected outcomes in real time. The modular architecture allows the framework to be extended with new perception methods, reasoning capabilities, or domain-specific knowledge without requiring complete system redesign. Consequently, it can accommodate task executions of varying complexity, from simple manipulation tasks to multi-step protocols involving complex object interactions. Through this integrated approach, the semantic execution tracing framework enables a new level of task-level documentation that supports both immediate reproducibility and long-term scientific analysis of robot-executed proce-Figure 1: Replaying NEEMs in the VRB.