Exploiting Redundancy for Reliability Analysis of Sensor Perception in Automated Driving Vehicles

A Review of Testing Object-Based Environment Perception for Safe Automated Driving

“…The mentioned testing approach exploits the redundancy of fusion inputs in the following way [31]. If a mean time between failures of 10 4 h can be demonstrated for two statistically independent inputs to a data fusion system, the fusion output has a mean time between failures of about 10 8 h .…”

“…set From undocumented search: Salay et al[27,28], Salay and Czarnecki[29], Czarnecki and Salay[30], Berk et al[31,32], Shalev-Shwartz et al [33], Amersbach and Winnner [5], Stellet et al [13], Brahmi et al [34], Cassel et al [35]; From undocumented and keyword search: Berk [36], Stellet et al [4], Salay et al [37], Ulbrich et al [11]; From keyword search: Borg et al [17], Willers et al [21], Burton et al [22], Taş et al [38], Kaprocki et al [39], Martin et al [40], Bai et al *, Johansson and Nilsson [52]*, Johansson et al [53]*, Bock et al [54]*, Rivero et al [55]*, Rasshofer et al [56]*, Fawcett [57]*, Cao and Huang [58], Czarnecki et al [59]*, Meyer [60]*, Guo [61]*, Salay and Czarnecki [62], Schwaiger et al [63], Arnez et al [64], Piazzoni et al [65], Weast [66], Klamann et al [67]*, Aravantinos and Schlicht [68]*, Breitenstein et al [69]* (29 publications). Iteration 2 Henne et al [70], Fleck et al [71], Spanfelner et al [72]*, Rahimi et al [73], Czarnecki [74]*, Feng et al [75]*, [76] (7 publications).…”

A Redundancy Metric Set within Possibility Theory for Multi-Sensor Systems

“…Several papers build upon the correlation-based feature selection to improve implementations and fasten the search for redundancies in sets of sources such as [ 35 , 36 ]. More recent applications of correlation as a redundancy measure can be found in [ 8 , 37 , 38 , 39 ]. Goswami et al [ 37 ] cluster features based on their redundancy determined using the Pearson’s correlation coefficient.…”

“…Goswami et al [ 37 ] cluster features based on their redundancy determined using the Pearson’s correlation coefficient. In [ 38 , 39 ], redundant features are eliminated based on correlation coefficients for applications in biology, whereas Berk et al [ 8 ] determine reliability and redundancy of sensors in an automated driving scenario. However, there has been some debate in the feature selection community about the appropriateness of using correlation-based metrics.…”

“…Information fusion aims at reducing uncertainties by aggregating information from multiple sensors or sources [ 3 , 4 ]. In addition to reducing uncertainty, redundant information allows a fusion system both to increase its robustness and to identify unreliable, drifting, or malfunctioning sensors [ 5 , 6 , 7 , 8 ]. Designing an information fusion system involves the decision of which sensors are to be fused at which stage in the information processing.…”