Ab­stract

For au­tonomous dri­ving to op­er­ate safely it is cru­cial to per­ceive sur­round­ing ob­jects cor­rectly. Not only de­tec­tion but also state es­ti­ma­tion (track) of a per­ceived ob­ject is ur­gent. The state is re­quired to en­able a safe mo­tion plan­ning, since it al­lows to pre­dict the fu­ture po­si­tion of an ob­ject. To in­clude only valid in­for­ma­tion, the state es­ti­ma­tions must be main­tained to de­ter­mine which track is ac­tive and which is not. Mostly, a sim­ple count-based ap­proach is used. For this, we pre­sent an in­ves­ti­ga­tion of two com­mon ap­proaches from non-co­op­er­a­tive track man­age­ment in com­par­i­son to two new man­age­ment strate­gies to main­tain tracks in a co­op­er­a­tive sce­nario. We eval­u­ate them using three sim­u­lated sce­nar­ios with a vary­ing rate of co­op­er­a­tive ve­hi­cles. A con­fi­dence-based ap­proach was able to in­crease the av­er­age pre­ci­sion by up to 9 per­cent­age points. “This is an Orig­i­nal​ Man­u­script of an ar­ti­cle pub­lished by De Gruyter​ in at - Au­toma­tisierung­stech­nik on 7th April 2023, avail­able at https://​doi.​org/​10.​1515/​auto-​2022-​0157"