Method and system for collaboratively detecting tangible bodies in an environment

Patent metadata

Field	Value
Publication number	US-12032385-B2
Application number	US-202117473859-A
Country	US
Kind code	B2
Filing date	Sep 13, 2021
Priority date	Sep 14, 2020
Publication date	Jul 9, 2024
Grant date	Jul 9, 2024

Abstract

A method is provided for collaboratively detecting tangible bodies in an environment employing a plurality of distance sensors with which respective carriers are equipped, including: for each sensor, acquiring a series of measurements of distance of a closest tangible body along a line of sight; applying an inverse model of the sensor on a local occupancy grid; and constructing a consolidated local occupancy grid via Bayesian fusion of the occupancy probabilities thus determined; and, computing occupancy probabilities of the cells of a global occupancy grid via Bayesian fusion of the occupancy probabilities of corresponding cells of at least certain of said consolidated local occupancy grids. A system for implementing such a method is also provided.

First claim

The invention claimed is:

1. A method for collaboratively detecting tangible bodies in an environment, this method employing a plurality of distance sensors with which respective carriers located in said environment are equipped, at least one of said carriers being mobile, each distance sensor having a position and orientation that are known with a known level of uncertainty, the method comprising the following steps:
step A) for each distance sensor, acquiring a series of measurements of distance of a closest tangible body along a line of sight; applying, to each said measurement of distance, an inverse model of the distance sensor on a local occupancy grid providing a discretized spatial representation of its field of view, to determine a probability of occupancy by a tangible body of a set of cells of said local occupancy grid; and constructing a consolidated local occupancy grid, each cell of which has an occupancy probability computed via Bayesian fusion of occupancy probabilities thus determined; and
step B) computing occupancy probabilities of cells of a global occupancy grid via Bayesian fusion of the occupancy probabilities of corresponding cells of at least certain of said consolidated local occupancy grids corresponding to each distance sensor with which separate carriers are equipped.

2. The method according to claim 1 , wherein step B) comprises the following substeps, for each cell of the global occupancy grid:
b1) identifying a subset of cells of each consolidated local occupancy grid meeting a condition of proximity with said cell of the global occupancy grid;
b2) computing a level of proximity of each cell thus identified with said cell of the global occupancy grid;
b3) for each local occupancy grid, computing a contribution to the occupancy probability via a weighted mean of the occupancy probabilities of the cells identified in substep b1), weighting coefficients being dependent on corresponding levels of proximity of each cell; and
b4) carrying out Bayesian fusion of the contributions to the occupancy probabilities computed in substep b3).

3. The method according to claim 2 , wherein substep b1) comprises identifying cells of each consolidated local occupancy grid that are located entirely or partially inside a region, of a given radius, centred on the cell of the global probability grid.

4. The method according to claim 3 , wherein said radius increases as the level of uncertainty in the position of the distance sensor corresponding to the consolidated local occupancy grid increases.

5. The method according to claim 2 , wherein the level of proximity of each cell computed in substep b2) is representative of a density of a probability that the centre of the identified cell coincides with the centre of the cell of the global occupancy grid.

6. The method according to claim 1 , wherein step A) is implemented by a plurality of local processing units, each associated with one carrier, and step B) is implemented by a central processing unit, the method also comprising transmitting the consolidated local occupancy grids from said distance sensors to said central processing unit.

7. The method according to claim 1 , comprising a prior step of determining a reference position and an orientation defining a reference global coordinate system, wherein step A) comprises, for each sensor, constructing said consolidated local occupancy grid aligned and located in the reference global coordinate system, the occupancy probabilities of said consolidated local occupancy grid being computed taking into account the level of uncertainty in the position and orientation of the sensor.

8. The method according to claim 7 , wherein a distance sensor has a position and an orientation that are set and known without uncertainty, these being taken as the reference position and as the orientation of the reference global coordinate system.

9. The method according to claim 7 , wherein:
a central processing unit determines the reference position and the orientation defining the reference global coordinate system and transmits them to a plurality of local processing units, each associated with one carrier;
step A) is implemented by one of said plurality of local processing units;
the consolidated local occupancy grids are transmitted by the local processing units to the central processing unit; and
step B) is implemented by said central processing unit.

10. The method according to claim 1 , wherein each said inverse model of the distance sensor is a discrete model, associating, with each cell of a corresponding local occupancy grid, and for each measurement of distance, a probability class chosen from inside a given set of finite cardinality, each said probability class being identified by an integer index; the Bayesian fusion operations being performed by means of integer computations performed on the integer indices of each probability class determined in said step b).

11. A system for collaboratively detecting tangible bodies, comprising:
a plurality of distance sensors with which respective carriers are equipped, at least one of said carriers being mobile;
a plurality of locating devices with which at least said one or more mobile carriers are equipped, said locating devices being suitable for determining a position and an orientation of said one or more mobile carriers, and their levels of uncertainty; and
a processing system configured or programmed to:
for each distance sensor, acquire a series of measurements of distance of a closest tangible body along a line of sight; apply, to each said measurement of distance, an inverse model of the distance sensor on a local occupancy grid providing a discretized spatial representation of its field of view, to determine a probability of occupancy by a tangible body of a set of cells of said local occupancy grid; construct a consolidated local occupancy grid, each cell of which has an occupancy probability computed via Bayesian fusion of occupancy probabilities thus determined; and
compute occupancy probabilities of cells of a global occupancy grid via Bayesian fusion of the occupancy probabilities of corresponding cells of at least certain of said consolidated local occupancy grids corresponding to each distance sensor with which separate carriers are equipped.

12. The system according to claim 11 , wherein the processing system comprises a plurality of local processing units each associated with one carrier, a central processing unit and a communication network connecting the central processing unit to each local processing unit, and wherein:
each local processing unit is configured or programmed to construct said consolidated local occupancy grid and to transmit it to the central processing unit; and
the central processing unit is configured or programmed to compute the occupancy probabilities of the cells of said global occupancy grid.

13. The system according to claim 12 , wherein the central processing unit is configured or programmed to, for each cell of the global occupancy grid:
identify a subset of cells of each consolidated local occupancy grid meeting a condition of proximity with said cell of the global occupancy grid; and
compute a level of proximity of each cell thus identified with said cell of the global occupancy grid;
for each local occupancy grid, compute an occupancy probability via a weighted mean of the occupancy probabilities of the identified cells, weighting coefficients being dependent on corresponding levels of proximity of each cell; and
carrying out Bayesian fusion of the probabilities thus computed.

14. The system according to claim 12 , wherein the central

Assignees

• Commissariat Energie Atomique

Inventors

• Pereira Da Silva Alex
• Leirens Sylvain
• Lesecq Suzanne
• Puschini Pascual Diego

Classifications

• G05D1/246

  using environment maps, e.g. simultaneous localisation and mapping [SLAM] · CPC title

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06F18/253

  of extracted features · CPC title

• G06V20/58

  Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads · CPC title

• G06F18/25

  Fusion techniques · CPC title