Traceability for Autonomous Systems: A Standardized Blockchain-Based Black Box

[EN] As autonomous systems become increasingly widespread in contemporary society, ensuring both their security and the safety of the people interacting with them has grown ever more critical. Autonomous systems must be designed around two fundamental principles: accountability and explainability. Much like airplane black boxes that capture critical flight data, these systems should incorporate secure modules that enable thorough postincident analysis. This article proposes a prototype that aims to enhance the existing black boxes for autonomous systems. Specifically, a standard for the storage of valuable information from autonomous systems is introduced, along with two implementations. First, BCubed is a library that stores data in blockchain following the structure defined by the proposed standard and allows for information retrieval. Second, ACOLYTE is an application that extracts the data and stores them employing BCubed. To evaluate the effectiveness of the application and its performance in a recognized testing environment, a series of experiments was conducted using a blockchain simulator. These experiments are grounded in the “Robot Inspection” test of the RoboCup@Home league. The results indicate that the information is accurately stored, and the metrics impacted by the application’s execution are slight increase in the percentage of CPU and memory usage by the autonomous system. In addition, a large language model was used to perform an objective analysis of the recorded data, and its response confirms the effectiveness of the solution in identifying the root causes of these anomalies.