System and method for controlling a mobile industrial robot using a probabilistic occupancy grid
A robot movable on a substrate is controlled on the basis of an occupancy grid of cells, where each cell is associated with an occupancy probability that some physical object is present in the cell. Occupancy-related measurements are obtained by RGB-D, radar or other sensing of an incident electromagnetic wave at an elevated point on the robot. From a measurement taken at an angle of incidence, an occupancy probability is assigned as follows: it is evaluated whether the measurement indicates that some physical object is present in the angle of incidence; a first predetermined model is selected if the evaluation is positive, and second predetermined model if the evaluation is negative; and a new occupancy probability for cells in the angle of incidence is determined on the basis of the measurement and the selected model.
1 . A method of controlling an industrial robot, which is movable on a substrate, comprising:
initializing an occupancy grid of cells which each represents a portion of the substrate and is associated with an occupancy probability that some physical object is present in the cell;
obtaining occupancy-related measurements using a plurality of measuring principles that include sensing an incident electromagnetic wave at an elevated point on the robot, the plurality of measuring principles including RGB-D sensing fused with radar;
assigning an occupancy probability on the basis of an obtained occupancy-related measurement at an angle of incidence; and
controlling the industrial robot on the basis of the occupancy grid,
wherein the assigning of the occupancy probability includes:
evaluating whether the measurement indicates that some physical object is present in the angle of incidence,
selecting a first predetermined model if the evaluation is positive and selecting a second predetermined model if the evaluation is negative, and
determining a new occupancy probability for cells in the angle of incidence on the basis of the measurement and in accordance with the selected model;
wherein:
the second predetermined model stipulates that the new occupancy probability in cells in a sensor frustum with a particular solid angle of incidence shall be in a positive relation to a local vertical thickness of the sensor frustum, and
the first predetermined model is adaptive depending on the measuring principle by which the measurement was obtained, such that the first predetermined model stipulates that:
the new occupancy probability in cells where the measurement indicates presence of some physical object shall be constant when the measurement is obtained by RGB-D sensing, and
the new occupancy probability in cells where the measurement indicates presence of some physical object shall be in a negative relation to a noisiness of the radar signal when the measurement is obtained by radar.
2 . The method of claim 1 , wherein the second model is adaptive in dependence of the at least one measuring principle by which the measurement was obtained.
3 . The method of claim 1 , wherein the measuring principles include one or more of:
optical, electromagnetic reflection, electromagnetic scattering, electromagnetic diffraction, lidar, RGB-D sensing, mm-wave radar, or ultra-wideband radar.
4 . The method of claim 1 , wherein assigning an occupancy probability includes:
merging the new occupancy probability and a pre-existing occupancy probability for the cell, in particular by applying a recursive rule such as Bayes' rule.
5 . The method of claim 4 , wherein the merging of the new occupancy probability and a pre-existing occupancy probability for the cell is performed by applying a recursive rule such as Bayes' rule.
6 . The method of claim 1 , wherein the second predetermined model stipulates that the new occupancy probability shall be linearly related to the local vertical thickness of the sensor frustum.
7 . The method of claim 1 , wherein:
the second predetermined model includes applying a pre-correction factor, which maintains the new occupancy probability in off-center cells closer to a neutral probability when the measuring principle includes radar.
8 . The method of claim 2 , wherein the stone measuring principles include one or more of:
optical, electromagnetic reflection, electromagnetic scattering, electromagnetic diffraction, lidar, RGB-D sensing, mm-wave radar, or ultra-wideband radar.
9 . The method of claim 2 , wherein assigning an occupancy probability includes:
merging the new occupancy probability and a pre-existing occupancy probability for the cell, in particular by applying a recursive rule such as Bayes' rule.
10 . A robot controller configured to control at least one industrial robot movable on a substrate, wherein the robot is controlled on the basis of an occupancy grid of cells which each represents a portion of the substrate and is associated with an occupancy probability that some physical object is present in the cell, the robot controller including:
an input interface for receiving occupancy-related measurements obtained by a plurality of measuring principles that include sensing an incident electromagnetic wave at an elevated point on the robot, the plurality of measuring principles including RGB-D sensing fused with radar;
processing circuitry configured to assign an occupancy probability on the basis of an obtained occupancy-related measurement at an angle of incidence, including:
evaluating whether the measurement indicates that some physical object is present in the angle of incidence,
selecting a first predetermined model if the evaluation is positive and selecting a second predetermined model if the evaluation is negative, and
determining a new occupancy probability for cells in the angle of incidence on the basis of the measurement and in accordance with the selected model; and
an output interface for supplying commands to the industrial robot;
wherein:
the second predetermined model stipulates that the new occupancy probability in cells in a sensor frustum with a particular solid angle of incidence shall be in a positive relation to a local vertical thickness of the sensor frustum, and
the first predetermined model is adaptive depending on the measuring principle by which the measurement was obtained, such that the first predetermined model stipulates that:
the new occupancy probability in cells where the measurement indicates presence of some physical object shall be constant when the measurement is obtained by RGB-D sensing, and
the new occupancy probability in cells where the measurement indicates presence of some physical object shall be in a negative relation to a noisiness of the radar signal when the measurement is obtained by radar.
11 . A computer program comprising instructions for causing a robot controller to perform a method including the following steps:
initializing an occupancy grid of cells which each represents a portion of the substrate and is associated with an occupancy probability that some physical object is present in the cell;
obtaining occupancy-related measurements using a plurality of measuring principles that include sensing an incident electromagnetic wave at an elevated point on the robot, the plurality of measuring principles including RGB-D sensing fused with radar;
assigning an occupancy probability on the basis of an obtained occupancy-related measurement at an angle of incidence; and
controlling the industrial robot on the basis of the occupancy grid,
wherein the assigning of the occupancy probability includes:
evaluating whether the measurement indicates that some physical object is present in the angle of incidence,
selecting a first predetermined model if the evaluation is positive and selecting a second predetermined model if the evaluation is negative, and
determining a new occupancy probability for cells in the angle of incidence on the basis of the measurement and in accordance with the selected model;
wherein:
the second predetermined model stipulates that the new occupancy probability in cells in a sensor frustum with a particular solid angle of incidence shall be in a positive relation to a local vertical thickness of the sensor frustum, and
the first predetermined model is adaptive depending on the measuring principle by which the measurement was obtained, such that the first predetermined model stipulates that:
the new occupancy probability in cells where the measurement indicates presence of some physical object shall be constant when the measurement is obtained RGB-D sensing, and
the new occupancy probability in cells where the measurement indicates presence of some physical object shall be in a negative relation to a noisiness of the radar signal when the measurement is obtained by radar.
12 . A data carrier storing a computer program for controlling a robot controller using the steps of:
initializing an occupancy grid of cells which each represents a portion of the substrate and is associated with an occupancy probability that some physical object is present in the cell;
obtaining occupancy-related measurements using a plurality of measuring principles that include sensing an incident electromagnetic wave at an elevated point on the robot, the plurality of measuring principles including RGB-D sensing fused with radar;
assigning an occupancy probability on the basis of an obtained occupancy-related measurement at an angle of incidence; and
controlling the industrial robot on the basis of the occupancy grid,
wherein the assigning of the occupancy probability includes:
evaluating whether the measurement indicates that some physical object is present in the angle of incidence,
selecting a first predetermined model if the evaluation is positive and selecting a second predetermined model if the evaluation is negative, and
determining a new occupancy probability for cells in the angle of incidence on the basis of the measurement and in accordance with the selected model;
wherein:
the second predetermined model stipulates that the new occupancy probability in cells in a sensor frustum with a particular solid angle of incidence shall be in a positive relation to a local vertical thickness of the sensor frustum, and
the first predetermined model is adaptive depending on the measuring principle by which the measurement was obtained, such that the first predetermined model stipulates that:
the new occupancy probability in cells where the measurement indicates presence of some physical object shall be constant when the measurement is obtained by RGB-D sensing,
the new occupancy probability in cells where the measurement indicates presence of some physical object shall be in a negative relation to a noisiness of the radar signal when the measurement is obtained by radar.