Functional system architectures towards fully automated driving

Taş, Ömer Şahin; Kuhnt, Florian; Zöllner, J. Marius; Stiller, Christoph · 2016 · OpenAlex-citations

DOI: 10.1109/ivs.2016.7535402

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This paper addresses the critical role of functional system architectures in achieving robust and reliable fully automated driving. While significant progress has been made in perception, prediction, and planning algorithms, the authors argue that system architecture design has received insufficient attention despite its massive influence on fault detection and overall vehicle reliability. Motivated by the need to cope with uncertainties such as sensor failures, measurement inaccuracies, and unexpected traffic situations, the study investigates existing architectures to derive requirements for future robust systems. The analysis focuses on SAE automation levels 3 through 5, where robustness is vital for safe operation and minimizing driver take-over requests. The authors categorize architectures into centralized and distributed types, noting that centralized designs suffer from high computational burdens and poor fault isolation, while distributed architectures offer better modularity but require secured communication protocols. The study reviews the architectural details of several state-of-the-art autonomous vehicles, including Boss (DARPA Urban Challenge winner), Bertha, AnnieWAY, A1 (Korean Autonomous Vehicle Competition winner), Shelley (autonomous race car), the Ulm University vehicle, and Audi’s Jack. The analysis evaluates these systems based on their handling of sensor fusion, failure monitoring, recovery mechanisms, and feedback loops between planning and control modules. Key findings reveal that vehicles capable of handling faults and exhibiting renowned performance share specific architectural traits. Successful systems predominantly utilize distributed architectures with redundant, complementary sensors covering the vehicle’s surroundings. They employ monitoring systems that supervise module health and trigger recovery instructions or degraded operation modes upon failure. Notably, robust architectures integrate uncertainty propagation into perception and planning, use feedback from controllers to motion planners, and allow for graceful degradation of software functions. For instance, Boss utilized a progress monitoring system for recovery, while Audi’s Jack used Bayesian networks to reason about discrepancies between map-based localization and environmental perception. Conversely, vehicles like Bertha and AnnieWAY lacked dedicated degraded operation modules, limiting their robustness. The significance of this work lies in its identification of essential requirements for future functional system architectures. The authors conclude that robustness is increasingly becoming a primary concern in highly automated vehicle development. They propose that future architectures should integrate existing solution approaches into a unified framework where every module employs metrics to quantify performance status. This would enable the system to switch into degraded operation modes, activate alternative data processing algorithms, and adapt trajectory planning based on perception guarantees. By ensuring safe but not overly conservative trajectories and allowing for algorithm adaptation during temporary sensor degradation, such architectures will significantly enhance the reliability required for fully automated driving.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• driverless ads