Sensor Object Plausibilization with Boids Flocking Algorithm

TL;DR

This paper introduces a novel approach using the Boids flocking algorithm to enhance the plausibility of sensor-based object tracking in driver assistance systems, improving target discrimination with minimal computational overhead.

Contribution

It applies the Boids flocking algorithm to model interactions between road users, improving object separation estimates in low-cost sensor environments.

Findings

Median separation improved from 2.4 m to 3 m

Bottom percentile separation increased from 1.85 m to 2.8 m

Effective with only 7 boids per traffic participant

Abstract

Driver assistance systems are increasingly becoming part of the standard equipment of vehicles and thus contribute to road safety. However, as they become more widespread, the requirements for cost efficiency are also increasing, and so few and inexpensive sensors are used in these systems. Especially in challenging situations, this leads to the fact that target discrimination cannot be ensured which in turn leads to a false reaction of the driver assistance system. Typically, the interaction between moving traffic participants is not modeled directly in the environmental model so that tracked objects can split, merge or disappear. The Boids flocking algorithm is used to model the interaction between road users on already tracked objects by applying the movement rules (separation, cohesion, alignment) on the boids. This facilitates the creation of semantic neighborhood information between road users. We show in a comprehensive simulation that with only 7 boids per traffic participant, the estimated median separatation between objects can improve from 2.4 m to 3 m for a ground truth of 3.7 m. The bottom percentile improves from 1.85 m to 2.8 m.