LMB Filter Based Tracking Allowing for Multiple Hypotheses in Object Reference Point Association

Abstract

Autonomous vehicles need precise knowledge on dynamic objects in their surroundings. Especially in urban areas with many objects and possible occlusions, an infrastructure system based on a multi-sensor setup can provide the required environment model for the vehicles. Previously, we have published a concept of object reference points (e.g. the corners of an object), which allows for generic sensor "plug and play" interfaces and relatively cheap sensors. This paper describes a novel method to additionally incorporate multiple hypotheses for fusing the measurements of the object reference points using an extension to the previously presented Labeled Multi-Bernoulli (LMB) filter. In contrast to the previous work, this approach improves the tracking quality in the cases where the correct association of the measurement and the object reference point is unknown. Furthermore, this paper identifies options based on physical models to sort out inconsistent and unfeasible associations at an early stage in order to keep the method computationally tractable for real-time applications. The method is evaluated on simulations as well as on real scenarios. In comparison to comparable methods, the proposed approach shows a considerable performance increase, especially the number of non-continuous tracks is decreased significantly.

Distributed Implementation of the Centralized Generalized Labeled Multi-Bernoulli Filter

“…k = T . Then, using the PM-δ-GLMB filter, the total computational complexity for each source hypothesis yields O(V T log T + T V s=1 (M (s) + 2P ) 4 ), if the Murty algorithm is used, and O(V T log T + T P 2 V s=1 M (s) ), if a Gibbs sampler is used. This means that the complexity of the fusion (first summand) is negligible compared to the cost of the sensor updates (second summand) in both cases, resulting in the complexities O(T V s=1 (M (s) + 2P ) 4 ) (Murty) and O(T P 2 V s=1 M (s) ) (Gibbs).…”

“…due to sensor limitations, large environments or occlusions. One example is the field of Cooperative Intelligent Transportation Systems, where environment models from distributed infrastructure sensors [3], [4] can help automated vehicles [5] and human drivers [6], especially in complex urban scenarios [7].…”

Handling Occlusions in Automated Driving Using a Multiaccess Edge Computing Server-Based Environment Model From Infrastructure Sensors

“…Further extensions and approximations to reduce the computational effort and to allow for a real-time application have been developed, like the Labeled Multi-Bernoulli (LMB) filter [40]. Our approach presented in this paper uses a centralized LMB Multi-Object Tracker (MOT) [18], [19]. However, the application would also allow for a distributed implementation of the Bayes-optimal GLMB filter between the MEC server and the sensor processing units (SPUs), as we show in [17].…”

A Two-Stage Feature Point Detection and Marking Approach Based on the Labeled Multi-Bernoulli Filter

“…Moratuwage, D. et al [ 31 ] presented a SLAM solution using an efficient variant of the δ-GLMB filter (δ-GLMB-SLAM) based on Gibbs sampling, which is computationally comparable to LMB-SLAM, yet more accurate and robust against sensor noise, measurement clutter, and feature detection uncertainty. Herrmann, M. et al [ 32 ] described a novel method to additionally incorporate multiple hypotheses for fusing the measurements of the object reference points using an extension to the previously presented Labeled Multi-Bernoulli (LMB) filter.…”