Driving assistance method and system

Patent metadata

Field	Value
Publication number	US-11364899-B2
Application number	US-201716617941-A
Country	US
Kind code	B2
Filing date	Jun 2, 2017
Priority date	Jun 2, 2017
Publication date	Jun 21, 2022
Grant date	Jun 21, 2022

Abstract

A driving assistance system includes a sensor set, a data storage device and an output device. The sensor set observes apparent states of each road user of the plurality of road users at successive time steps, the data processor assigns a behavioral model stored in the data storage device, the data processor calculating a new maneuver distribution that is a probability distribution over a finite plurality of alternative maneuvers and a new state distribution that is a probability distribution of possible states for each alternative maneuver of the finite plurality of alternative maneuvers. The data processor determines a risk of collision of the road vehicle with another road user, based on the new maneuver and state distributions of the target road user, and outputs one or more of a driver warning signal and executes an avoidance action if the risk of collision exceeds a predetermined threshold.

First claim

The invention claimed is:

1. A driving assistance method for a road vehicle, the
driving assistance method comprising the steps of:
observing, in a traffic scene including the road vehicle among a plurality of road users, apparent states of each road user of the plurality of road users at successive time steps;
assigning a behavioral model to a target road user of the plurality of road users;
calculating, for the target road user, at a new time step, based on the apparent states that have been observed, a new maneuver distribution that is a probability distribution over a finite plurality of alternative maneuvers, and a new state distribution that is a probability distribution over possible states for each alternative maneuver of the finite plurality of alternative maneuvers, wherein:

a probability of a selected maneuver in the new maneuver distribution is calculated as proportional to a sum, over all previous time step's maneuvers, of a product, for each previous time step's maneuver that is a possible maneuver of the target road user in a previous time step, of a first term that is a previously calculated probability of the previous time step's maneuver, a second term that is a probability of the selected maneuver in the new time step from a probability distribution over possible states for the previous time step's maneuver based on the behavioral model assigned to the target road user, and a third term that is a probability of the apparent state observed at the new time step calculated for the selected maneuver and the probability distribution over possible states for the previous time step's maneuver based on a set of motion parameters associated to the selected maneuver; and
the new distribution over possible states for each alternative maneuver is calculated as the sum, over all previous time step's maneuvers, of the product, for each previous time step's maneuver, of at least a state distribution obtained by propagating the probability distribution of possible states for the previous time step's maneuver using the motion parameters associated to each alternative maneuver and a fourth term proportional to the product of the first, second and third terms;

determining a risk of collision of the road vehicle with another road user of the plurality of road users, based on the new maneuver and state distributions of the target road user; and
executing an avoidance action if the risk of collision exceeds a predetermined threshold.

2. The driving assistance method of claim 1 , wherein the third term is calculated by applying a set of motion parameters associated to the selected maneuver to the probability distribution over possible states for the previous time step's maneuver so as to obtain a dynamics-based predicted probability distribution for the new time step which is then compared with the apparent state observed at the new time step.

3. The driving assistance method of claim 2 , wherein the set of motion parameters associated to the selected maneuver is applied using at least a prediction step of an Extended Kalman Filter algorithm.

4. The driving assistance method of claim 1 , wherein the behavioral model assigned to the target road user takes the form of a cost function for calculating a cost of a state of the target road user, with at least one dynamic component for taking into account a state of a road user other than the target road user, and the second term is calculated by:
sampling a plurality of possible states from the probability distribution among possible states of the target road user for the previous time step's maneuver;
propagating each possible state sampled for the target road user, over a plurality of subsequent time steps, according to the set of motion parameters associated to each maneuver of the finite plurality of alternative maneuvers, to obtain an alternative sequence of prospective states at the plurality of subsequent time steps for each possible state sampled for each maneuver of the finite plurality of alternative maneuvers and the target road user;
sampling at least one possible state and maneuver, from a state and a maneuver distribution of at least a road user, of the plurality of road users, other than the target road user, for the previous time step;
propagating the at least one possible state sampled for the at least one road user other than the target road user, according to a set of motion parameters associated to the at least one maneuver sampled for the at least one road user other than the target road user, to obtain at least one sequence of prospective states at the plurality of subsequent time steps for the at least one road user other than the target road user;
estimating a cost of each prospective state of each alternative sequence of prospective states of the target road user, according to the behavioral model assigned to the target road user, taking into account the at least one prospective state of the at least one road user other than the target road user at the same subsequent time step;
aggregating the costs of the prospective states of each alternative subsequent sequence of prospective states of the target road user to obtain an aggregated cost for each alternative subsequent sequence of prospective states of the target road user;
averaging the aggregated costs of the alternative sequence of prospective states of the target road user for each maneuver of the finite plurality of alternative maneuvers to obtain an average aggregated cost of each maneuver of the finite plurality of alternative maneuvers; and
subtracting from one the ratio of the average aggregated cost of the selected maneuver to the sum of the average aggregated costs of the finite plurality of alternative maneuvers.

5. The driving assistance method of claim 4 , wherein the at least one dynamic component comprises one or more of a time-headway and a time-to-collision between the target road user and another road user of the plurality of road users.

6. The driving assistance method of claim 4 , wherein, to assign a behavioral model to the target road user of the plurality of road users, an aggregated cost of successive observed states of the target road user is calculated for each behavioral model of a finite plurality of alternative behavioral models, and the behavioral model with the lowest aggregated cost is selected.

7. The driving assistance method of claim 1 , wherein, to obtain a state distribution by propagating the probability distribution of possible states for each previous time step's maneuver when calculating the new distribution of possible states for each possible maneuver, the set of motion parameters associated to each possible maneuver is applied using an Extended Kalman Filter algorithm, wherein the apparent state of the target road user at the new time step is used in an updating step.

8. The driving assistance method according to claim 1 , wherein the behavioral model assigned to the target road user is selected from among a finite plurality of behavioral models.

9. The driving assistance method of claim 8 , wherein the finite plurality of behavioral models is learned from observed road user behavior using a machine learning algorithm.

10. The driving assistance method of claim 9 , wherein the traffic scene comprises a multi-lane road, and the finite plurality of alternative maneuvers comprises a lane-keeping and a lane-changing maneuver.

11. A driving assistance system for a road vehicle, the driving assistance system comprising:
a sensor set configured to observe a plurality of successive states for each road user of a plurality of road users in a traffic scene including the road vehicle among the plurality of road users;
a data storage device fo

Assignees

• Toyota Motor Europe
• Institut National De Rech En Informatique Et En Automatique Inria
• Inst Nat Sciences Appliquees Lyon
• Institut Nat Des Sciences Applquees De Lyon

Inventors

• Vignard Nicolas
• Sierra Gonzalez David
• Laugier Christian
• Dibangoye Jilles Steeve
• Romero-Cano Victor

Classifications

• G05B13/0265

  the criterion being a learning criterion · CPC title

• B60W50/14 Primary

  Means for informing the driver, warning the driver or prompting a driver intervention · CPC title

• B60W2554/00

  Input parameters relating to objects · CPC title

• B60W2554/4049

  Relationship among other objects, e.g. converging dynamic objects · CPC title

• B60W60/00276

  for two or more other traffic participants · CPC title