SecForCARs logo

Self-Assessment in Tracking Algorithms

Ulm University

Self-assessment is a key to safety and security in automated driving. In order to design safer as well as more robust and secure automated driving functions, the goal is to self-assess the performance of each module in a whole automated driving system. One crucial component in automated driving systems is the tracking of surrounding objects, where the Kalman filter is the most fundamental tracking algorithm.


Inaccurate, poor and, especially, malicious sensor data from attackers, which are authenticated as correct but semantically wrong, can lead to serious problems in sensor data post-processing. The self-assessment in tracking modules should be able to detect these issues, send out warnings for further processing steps or directly do some adaption of the tracking filter assumptions. In this work, we present a novel approach to obtain a self-assessment in Kalman filtering for single-object tracking in clutter using subjective logic. Subjective logic is a mathematical theory that explicitly models statistical uncertainty similar to the Dempster-Shafer theory. Thus, our approach features reliability measures that explicitly include statistical uncertainties. This additional statistical uncertainty measures can be particularly beneficial if the number of samples is strictly limited, e.g., due to a fast-changing environment as we often observe in automated driving. Our proposed self-assessment method is able to online estimate the tracking algorithm’s performance in order to ensure the safety and security of the whole automated driving system.

Corresponding Publication:
T. Griebel, J. Müller, M. Buchholz and K. Dietmayer, “Kalman Filter Meets Subjective Logic: A Self-Assessing Kalman Filter Using Subjective Logic,” 2020 IEEE 23rd International Conference on Information Fusion (FUSION), 2020, pp. 1-8.
DOI: FUSION45008.2020.9190520