System and method for radar cross traffic tracking and maneuver risk estimation

Patent metadata

Field	Value
Publication number	US-11155258-B2
Application number	US-201916362889-A
Country	US
Kind code	B2
Filing date	Mar 25, 2019
Priority date	Mar 25, 2019
Publication date	Oct 26, 2021
Grant date	Oct 26, 2021

Abstract

A risk maneuver assessment system and method to generate a perception of an environment of a vehicle and a behavior decision making model for the vehicle; a sensor system configured to provide the sensor input in the environment for filtering target objects; one or more modules configured to map and track target objects to make a candidate detection from multiple candidate detections of a true candidate detection as the tracked target object; apply a Markov Random Field (MRF) algorithm for recognizing a current situation of the vehicle and predict a risk of executing a planned vehicle maneuver at the true detection of the dynamically tracked target; apply mapping functions to sensed data of the environment for configuring a machine learning model of decision making behavior of the vehicle; and apply adaptive threshold to cells of an occupancy grid for representing an area of tracking of objects within the vehicle environment.

First claim

What is claimed is:

1. A risk maneuver assessment system for planning maneuvers with uncertainties of a vehicle, comprising:
a controller with a processor programmed to generate a perception of an environment of the vehicle and a behavior decision making model for the vehicle including performing a calculation upon a sensor input to provide, as an output an action risk mapping and at least one target object tracking for different areas within the environment of the vehicle;
a sensor system configured to provide the sensor input to the processor for providing an area in the environment of the vehicle for filtering target objects;
one or more modules configured to, by the processor, map and track the target objects to make a candidate detection from multiple candidate detections of a true candidate detection as the tracked target object;
one or more first modules configured to, by the processor, apply a Markov Random Field (MRF) algorithm for recognizing a current situation of the vehicle in the environment and predict a risk of executing a planned vehicle maneuver at the true detection of the dynamically tracked target;
one or more second modules configured to, by the processor, apply mapping functions to sensed data of the environment for configuring a machine learning model of decision making behavior of the vehicle;
one or more third modules configured to, by the processor, apply adaptive threshold to cells of an occupancy grid configured for representing an area of tracking of objects within the vehicle environment;
the one or more of the second and/or third modules further configured to select as the true detection at least one of the candidate detections that is within a radius for the target, and the candidate detection that is closest to a first known mapped pathway;
the one or more of the second and/or third modules further configured to select as the true detection, the candidate that indicates a position and velocity that are consistent with a target traveling on a second known travel mapped pathway, and to select as a false detection, the candidate that indicates the position that is outside the second known pathway or the velocity is not consistent with the target traveling; and
a fourth module configured to compute, by a gating operation, a distance metric from a last position of the tracked target object to a predicted position less than a threshold distance related to one or more of the candidate detections.

2. The system of claim 1 , further comprising:
the one or more first modules programmed to generate a Markov Random Field (MRF) to recognize the current situation.

3. The system of claim 1 , further comprising:
one or more fifth modules configured to, by the processor, apply the Markov Random Field (MRF) algorithm representing the tracked target object in one or more cells of the occupancy grid by:
calculating an object measurement density for each tracked target object represented in the one or more cells of the occupancy grid;
spreading the density over a window comprising a set of cells of the occupancy grid represented by the tracked target object; and
spreading velocities over the window comprising a same set of cells of the occupancy grid represented by the tracked target object.

4. The system of claim 3 , further comprising:
one or more sixth modules configured to, by the processor, apply mapping functions to the sensed data of the environment for configuring the machine learning (ML) model of decision making behavior of the vehicle by an action risk assessment model trained using semi-supervised machine learning techniques by on-line and off-line training for mapping the function to the candidate action to determine with risk factors a learned drivable path.

5. The system of claim 4 , further comprising:
a seventh module configured to, by a processor, perform in the off-line training of the ML model comprising:
collecting labels, co-collecting occupancy velocity grids, extracting features from the occupancy grids, and applying at least support vector machine (SVM) techniques for recognizing class patterns of the candidate actions to determine with risk factors the learned drivable path.

6. The system of claim 5 , further comprising an eighth module configured to, by the processor, apply the adaptive threshold to cells of the occupancy grid configured for representing the area of tracking of objects within the vehicle environment comprising:
a ninth module configured to, by the processor, compute by an adaptive threshold occupancy density, the likelihood that the candidate action is available for the target tracked object based on the computed density distribution, and select the candidate action that has a highest probability of being available; and
a tenth module configured to, by the processor, compute a clustering for velocity clusters for a set of candidate actions to select the target tracked object that indicates a position that is consistent with the learned drivable path.

7. The system of claim 5 , wherein the ML model is trained using reinforcement learning techniques using a data set of past collected labels and sensor data of drivable paths and wherein the eight module is configured to select the candidate action that will likely contribute to one of the drivable paths wherein the sensor data at least comprises one of: radar, acoustic, lidar or image sensor data.

8. A vehicle, comprising:
a sensor detection sensing device including one or more of a set comprising: a radar, acoustic, lidar and image sensing device;
a risk maneuver assessment system for assessing one or more uncertainty factors in planned maneuvers; and a plurality of modules configured to, by a processor, generate a perception of an environment of the vehicle and a output target output for tracking different areas within the environment; the plurality of modules comprising:
one or more modules configured to, by a processor, map and track target objects to make a candidate detection from multiple candidate detections of a true candidate detection as the tracked target object;
one or more modules configured to, by the processor, apply a Markov Random Field (MRF) algorithm for recognizing a current situation of the vehicle in the environment and for predicting a risk of executing a planned vehicle maneuver at the true detection of the dynamically tracked target;
one or more modules configured to, by the processor, apply mapping functions to sensed data of the environment for configuring a machine learning model of decision making behavior of the vehicle;
one or more modules configured to, by the processor, apply adaptive threshold to cells of an occupancy grid configured for representing areas of tracking of objects within the environment;
a controller with a processor configured to generate control commands in accordance with modeling of the decision making behavior and the perception of the environment of the vehicle for planned vehicle maneuvers;
a first module configured to select, as the true detection, the candidate detection that is within a radius for the target;
a second module configured to select, as the true detection, the candidate detection that is closest to a first known mapped pathway;
a third module configured to select the true detection, the candidate that indicates a position and velocity that are consistent with a target traveling on a second known travel mapped pathway, and to select a false detection, the candidate that indicates the position that is outside the second known pathway or the velocity is not consistent with the target traveling; and
a fourth module configured to compute, by a gating operation, a distance metric from the last position of the tracked target object to a predicted position less

Assignees

• Gm Global Tech Operations Llc

Inventors

• Bush Lawrence A
• Bajwa Manpreet S
• Zeng Shuqing
• Sprague Rickie A

Classifications

• G06N7/01

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

• B60W2554/4045

  Intention, e.g. lane change or imminent movement · CPC title

• B60W60/0016

  of the vehicle or its occupants · CPC title

• G01S13/931

  of land vehicles · CPC title

• G01S13/86

  Combinations of radar systems with non-radar systems, e.g. sonar, direction finder · CPC title