A random finite set approach for dynamic occupancy grid maps with real-time application

Abstract

Grid mapping is a well established approach for environment perception in robotic and automotive applications. Early work suggests estimating the occupancy state of each grid cell in a robot's environment using a Bayesian filter to recursively combine new measurements with the current posterior state estimate of each grid cell. This filter is often referred to as binary Bayes filter (BBF). A basic assumption of classical occupancy grid maps is a stationary environment. Recent publications describe bottom-up approaches using particles to represent the dynamic state of a grid cell and outline prediction-update recursions in a heuristic manner. This paper defines the state of multiple grid cells as a random finite set, which allows to model the environment as a stochastic, dynamic system with multiple obstacles, observed by a stochastic measurement system. It motivates an original filter called the probability hypothesis density / multi-instance Bernoulli (PHD/MIB) filter in a top-down manner. The paper presents a real-time application serving as a fusion layer for laser and radar sensor data and describes in detail a highly efficient parallel particle filter implementation. A quantitative evaluation shows that parameters of the stochastic process model affect the filter results as theoretically expected and that appropriate process and observation models provide consistent state estimation results.

Long-Term Occupancy Grid Prediction Using Recurrent Neural Networks

“…More recently, several works [2], [3], [4], [5] have extended the static grid map to dynamic environments, where the velocity has been introduced to the cell state. The advantage of dynamic occupancy grid maps is the ability to represent arbitrary shaped objects, while there is no need for an explicit object detection and data association.…”

“…Although the complete model is trained with supervision, we propose a pipeline for automatic label generation. To that end, we collect large amount of laser measurements that are pre-processed and then labeled with the support of existing algorithms [5], [7], [8].…”

“…However, the static environment assumption does not hold due to moving road users, such as pedestrians, vehicles and cyclists. Several works [2], [3], [4], [5] have proposed to extend static occupancy grid maps to dynamic occupancy grid maps, mainly based on particle filter approaches. Danescu et al [2] suggest to estimate position and velocity as continuous distribution using particles that can freely move between cells.…”

“…This idea has been further explored by Tanzmeister et al [4] and Nègre et al [3] by representing only the dynamic environment with particles. More recently, Nuss et al [5] defined the dynamic state estimation as a random finite set problem and implemented particles in the Dempster-Shafer domain. Unlike the prior work, we aim to estimate the occupancy and the velocity, based on recurrent neural networks.…”

Dynamic Hilbert Maps: Real-Time Occupancy Predictions in Changing Environments

“…Most of the early research studies on dynamic occupancy mapping fall under the first category in which the dynamic objects are The authors are with the 1 School of Computer Science, at The University of Sydney, Australia, 2 Toyota Research Institute, USA, and 3 NVIDIA Research, USA. Emails: { vitor.guizilini;ransalu.senanayake;fabio.ramos }@sydney.edu.au treated as spurious data and remove them to build a robust static map [5], [6], [7], [8]. In the second category, an occupancy pattern is obtained over a long period [5], [9], [10], [11] and such a map can later be used for global path planning.…”