Device for planning a path and/or trajectory for a motor vehicle

1. A device for planning a path and/or a trajectory for a motor vehicle, comprising:

processing circuitry configured to

receive training data corresponding to motor vehicle usage,

group the training data based on a type of maneuver of the motor vehicle,

extract characteristics associated with each group of training data,

train a recurrent neural network (RNN) using the extracted characteristics and a trajectory of the motor vehicle associated with each group of training data when the motor vehicle is stationary,

receive a sequence of input variables, the sequence of input variables corresponding to the extracted characteristics,

generate, using the trained RNN, a set of commands for the motor vehicle based on the sequence of input variables received; and

control the motor vehicle along a trajectory according to the set of commands,

the RNN including a connectionist temporal classification function to generate a set of compact representations of the sequence of input variables.

2. The device as claimed in claim 1 , wherein the processing circuitry is further configured to define a surjection of a set of inputs composed of different sequences of input variables of variable lengths toward a set of outputs composed of different sequences of output variables of variable lengths.

3. The device as claimed in claim 2 , wherein the set of inputs comprises a sequence of temporal input variables and/or a sequence of metric input variables.

4. The device as claimed in claim 2 , wherein the set of inputs comprises at least one sequence of input variables chosen from among a sequence of raw variables obtained from sensors, a sequence of variables for merged representation of an external environment of the vehicle, a sequence of variables for representation of internal parameters of the motor vehicle, a sequence of variables for representation of behaviors of an agent and a sequence of variables for representation of intentions of an agent.

5. The device as claimed in claim 2 , wherein the set of outputs comprises a sequence of temporal output variables and/or a sequence of metric output variables.

6. The device as claimed in claim 2 , wherein the set of outputs comprises at least an output sequence chosen from among a sequence of pairs of coordinates of route points, a sequence of abscissas of route points and a sequence of ordinates of route points.

7. The device as claimed in claim 1 , wherein the processing circuitry is configured to execute temporal classification of the sequence of input variables via a hidden Markov chain model.

8. The device as claimed in claim 1 , wherein the recurrent neural network is a long short-term memory recurrent neural network.

9. A method for planning a path and/or a trajectory for a motor vehicle, comprising:

receiving, via processing circuitry, training data corresponding to motor vehicle usage;

grouping, via the processing circuitry, the training data based on a type of maneuver of the motor vehicle;

extracting, via the processing circuitry, characteristics associated with each group of training data;

training, via the processing circuitry, a recurrent neural network (RNN) using the extracted characteristics and a trajectory of the motor vehicle associated with each group of training data when the motor vehicle is stationary;

receiving, via the processing circuitry, a sequence of input variables;

generating, via the processing circuitry, a set of commands for the motor vehicle using the RNN based on the sequence of input variables; and

controlling, via the processing circuitry, the motor vehicle along a trajectory according to the set of commands,

the RNN using a connectionist temporal classification function to generate a set of compact representations of the sequence of input variables.

10. The method as claimed in claim 9 , wherein the grouping the training data further comprising grouping a set of sequences of output variables corresponding, respectively, to the same maneuvers in a first level grouping.

11. The method as claimed in claim 10 , wherein, the grouping the training data further comprising grouping a set of sequences of variables corresponding, respectively, to maneuvers having substantially equal implementation periods and/or substantially equal implementation distances, in a second-level grouping.

12. The method as claimed in claim 9 , further comprising:

determining, via the processing circuitry, an index and at least one temporal modulation factor;

determining, via the processing circuitry and in a look-up table, a path associated with the index; and

temporally modulating, via the processing circuitry, the path by the temporal modulation factor transmitted.

13. The method as claimed in claim 9 , further comprising determining, via the processing circuitry, a temporal modulation factor in a longitudinal direction and/or a temporal modulation factor in a lateral direction.

14. A non-transitory computer readable medium storing a program that, when executed by a computer, causes the computer to execute a method, the method comprising:

receiving training data corresponding to motor vehicle usage;

grouping the training data based on a type of maneuver of the motor vehicle;

extracting characteristics associated with each group of training data;

training a recurrent neural network (RNN) using the extracted characteristics and a trajectory of the motor vehicle associated with each group of training data when the motor vehicle is stationary;

receiving a sequence of input variables;

generating a set of commands for the motor vehicle using the RNN based on the sequence of input variables; and

controlling the motor vehicle along a trajectory according to the set of commands,

the RNN using a connectionist temporal classification function to generate a set of compact representations of the sequence of input variables.