Modular Fault Diagnosis Framework for Complex Autonomous Driving Systems

TL;DR

The concept of a modular fault diagnosis framework for AD systems is presented, which suggests modular state monitoring and diagnosis elements, together with a state- and dependency-aware aggregation method and a classification scheme allows for the categorization of the fault diagnosis modules.

Abstract

Fault diagnosis is crucial for complex autonomous mobile systems, especially for modern-day autonomous driving (AD). Different actors, numerous use cases, and complex heterogeneous components motivate a fault diagnosis of the system and overall system integrity. AD systems are composed of many heterogeneous components, each with different functionality and possibly using a different algorithm (e.g., rule-based vs. AI components). In addition, these components are subject to the vehicle's driving state and are highly dependent. This paper, therefore, faces this problem by presenting the concept of a modular fault diagnosis framework for AD systems. The concept suggests modular state monitoring and diagnosis elements, together with a state- and dependency-aware aggregation method. Our proposed classification scheme allows for the categorization of the fault diagnosis modules. The concept is implemented on AD shuttle buses and evaluated to demonstrate its capabilities.

Figures (6)

• Figure 1: The modular diagnostic framework schematized here contributes to a system-wide diagnosis by considering the diagnostic states of the submodules (e.g. "3 of 3 Okay") of each component (green boxes, e.g. controller area network or "CAN") and their dependencies (orange arrows). The framework includes the current driving state of the AD vehicle, enabling the exclusion of irrelevant components from the diagnosis at a given time.
• Figure 2: A general classification scheme for faults in AD, focusing on fault detection methods. The taxonomy is based on fault diagnosis location, the type of information processed, and the data flow. At the bottom, examples of the different categories are presented. The examples can roughly be grouped into general components and data stream diagnosis techniques.
• Figure 3: The dependencies of the aggregated fault diagnosis groups, corresponding to the high-level system parts.
• Figure 4: Detailed view of all aggregated fault diagnosis groups corresponding to the system's parts, displayed in the human-machine interface of the shuttles. Here, the Localization group is in an $\textit{ERROR}$ state, resulting in dependent groups to be in the $\textit{IGNORE}$ state, depicted in grey with a label showing that they are "Ignored".
• Figure 5: The dependency graph of the fault diagnosis modules with their aggregated groups (grey), representing high-level system parts, of our AD shuttles. Dependencies between fault diagnosis modules represent dependencies of the components of the AD system. The driving state of the vehicle (Active) is depicted in green. This state is set by the Mission component and influences the behaviour of the Planning and Execution diagnosis modules.