Conditional Monte Carlo Dense Occupancy Tracker

Abstract

Proper modeling of dynamic environments is a core task in the field of intelligent vehicles. The most common approaches involve the modeling of moving objects, through Detection And Tracking of Moving Objects (DATMO) methods. An alternative to a classic object model framework is the occupancy grid filtering domain. Instead of segmenting the scene into objects and track them, the environment is represented as a regular grid of occupancy, in which spatial occupancy is tracked at a sub-object level. In this paper, we present the Conditional Monte Carlo Dense Occupancy Tracker, a generic spatial occupancy tracker, which infers dynamics of the scene through an hybrid representation of the environment, consisting of static occupancy, dynamic occupancy, empty spaces and unknown areas. This differentiation enables the use of state specific models (classic occupancy grid for motion-less components, set of moving particles for dynamic occupancy) as well as proper confidence estimation and management of data-less areas. The approach leads to a compact model that drastically improves the accuracy of the results and the global efficiency in comparison to previous methods.

Semantic Grid Estimation with a Hybrid Bayesian and Deep Neural Network Approach

“…A widely used method is to construct occupancy grids without having to recognize the objects in the surroundings. Occupancy grids are spatial 2D maps of the environment which also model regions containing moving objects [1], [2], [3]. The cells of these grids contain the probability of the state of a cell.…”

Natural vision based method for predicting pedestrian behaviour in urban environments

“…Future work will deal with the application of this framework from the ego-perspective of an autonomous car. Algorithms like the CMC-DOT [20] can be used for estimating the occupancy grid and positions of dynamic obstacles in real time. This can then be used as an input for our framework.…”

Validation of Perception and Decision-Making Systems for Autonomous Driving via Statistical Model Checking

“…The Conditional Monte Carlo Dense Occupancy Tracker (CMCDOT) Framework is a perception system, based on environment representation through probabilistic occupancy grids, a dense and generic representation [16], [17], and Bayesian fusion, filtering and inference [18].…”

“…While most of risk estimation methods consist in detecting and tracking dynamic objects in the scene [21], [22], the risk being then estimated through a Time to Collision (TTC) approach by projecting object trajectories to the future [23], [24], the grid-based approach used in the CMCDOT framework [18] instead directly computes estimations of the position in the near future of every static and dynamic part of the grid, as well as the trajectory of the vehicle. These estimations are iteratively computed over short time periods, until a potential collision is detected, in which case a TTC is associated to the cell from which the colliding element came from (Fig.…”