Method and apparatus for generating route planning model, and storage medium

1. A method for generating a target route planning model, comprising:

obtaining, by a computing device in a vehicle, a target route data set associated with a target site; and

in response to inputting target route data in the target route data set into a first route planning model, determining by the computing device a first error between a predicted trajectory outputted by the first route planning model and an actual trajectory corresponding to the target route data;

updating by the computing device a parameter of the first route planning model until the first error is smaller than a first threshold; and

determining by the computing device the updated first route planning model as the target route planning model of the target site, wherein the target route planning model is configured to output a trajectory of the vehicle for autonomous parking in the target site;

wherein the first route planning model is pre-trained based on a set of historical route data, the set of historical route data being associated with a plurality of sites different from the target site and comprising a plurality of first historical route data;

wherein the first route planning model is determined by:

in response to inputting the plurality of first historical route data into a first route planning sub-model, determining a second error between a predicted trajectory outputted by the first route planning sub-model and an actual trajectory corresponding to each first historical route data;

updating a parameter of the first route planning sub-model in a first gradient descent way until the second error is smaller than a second threshold; and

determining the updated first route planning sub-model as a the first route planning model;

wherein the method further comprises:

obtaining current route data of the target site by at least one sensor of the vehicle;

determining a trajectory for autonomous parking of the vehicle in the target site by inputting the current route data to the target route planning model; and

controlling steers, throttles, and brakes of the vehicle based on the determined trajectory and parking the vehicle in a predetermined parking space of the target site.

2. The method of claim 1 , wherein the first error is indicated by a target loss function corresponding to the target site, and the second error is indicated by a sum of site loss functions corresponding to the plurality of sites.

3. The method of claim 1 , wherein the set of historical route data further comprises a plurality of second historical route data, and the first route planning sub-model is determined by:

in response to inputting the plurality of second historical route data into an initial route planning model, determining a third error between a predicted trajectory outputted by the initial route planning model and an actual trajectory corresponding to each second historical route data; and

updating a parameter of the initial route planning model in a second gradient descent way until the third error is smaller than a third threshold; and

determining the updated initial route planning model as the first route planning sub-model.

4. The method of claim 1 , wherein obtaining the target route data set comprises:

obtaining target route data associated with the target site; and

obtaining the target route data set based on the target route data and a route data translation model, the target route data set comprising at least the target route data.

5. The method of claim 4 , wherein the route data translation model comprises a set of route data translation sub-models, each of the set of route data translation sub-models corresponds to a different environmental condition, and obtaining the target route data set further comprises:

determining a target environmental condition where the target route data is located;

determining one route data translation sub-model from the set of route data translation sub-models corresponding to the target environmental condition based on the target environmental condition; and

obtaining the target route data set based on the target route data and the determined route data translation sub-model.

6. An apparatus for generating a target route planning model, included in a vehicle, the apparatus comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor,

wherein the at least one processor is configured to:

obtain a target route data set associated with a target site;

in response to inputting target route data in the target route data set into a first route planning model, determine a first error between a predicted trajectory outputted by the first route planning model and an actual trajectory corresponding to the target route data;

update a parameter of the first route planning model until the first error is smaller than a first threshold; and

determine the updated first route planning model as the target route planning model of the target site, wherein the target route planning model is configured to output a trajectory of the vehicle for autonomous parking in the target site;

wherein the first route planning model is pre-trained based on a set of historical route data, the set of historical route data being associated with a plurality of sites different from the target site and comprising a plurality of first historical route data;

wherein the first route planning model is determined by:

in response to inputting the plurality of first historical route data into a first route planning sub-model, determining a second error between a predicted trajectory outputted by the first route planning sub-model and an actual trajectory corresponding to each first historical route data;

updating a parameter of the first route planning sub-model in a first gradient descent way until the second error is smaller than a second threshold; and

determining the updated first route planning sub-model as the first route planning model;

wherein the at least one processor is further configured to:

obtain current route data of the target site by at least one sensor of the vehicle;

determine a trajectory for autonomous parking of the vehicle in the target site by inputting the current route data to the target route planning model; and

control steers, throttles, and brakes of the vehicle based on the determined trajectory and park the vehicle in a predetermined parking space of the target site.

7. The apparatus of claim 6 , wherein the first error is indicated by a target loss function corresponding to the target site, and the second error is indicated by a sum of target loss functions corresponding to the plurality of sites.

8. The apparatus of claim 6 , wherein the set of historical route data further comprises a plurality of second historical route data, and the first route planning sub-model is determined by:

in response to inputting the plurality of second historical route data into an initial route planning model, determining a third error between a predicted trajectory outputted by the initial route planning model and an actual trajectory corresponding to each second historical route data; and

update a parameter of the initial route planning model in a second gradient descent way until the third error is smaller than a third threshold; and

determining the updated initial route planning model as the first route planning sub-model.

9. The apparatus of claim 6 , wherein the at least one processor is further configured to:

obtain target route data associated with the target site; and

obtain the target route data set based on the target route data and a route data translation model, the target route data set comprising at least the target route data.

10. The apparatus of claim 9 , wherein the route data translation model comprises a set of route data translation sub-models, each of the set of route data translation sub-models corresponds to a different environmental condition, and the at least one processor is further configured to:

determine a target environmental condition where the target route data is located; and

determine one route data translation sub-model from the set of route data translation sub-models corresponding to the target environmental condition based on the target environmental condition;

obtain the target route data set based on the target route data and the determined route data translation sub-model.

11. A non-transitory computer readable storage medium having computer instructions stored thereon, wherein the computer instructions are configured to cause a computer to execute a method for generating a target route planning model, the method comprising:

obtaining a target route data set associated with a target site; and

in response to inputting target route data in the target route data set into a first route planning model, determining a first error between a predicted trajectory outputted by the first route planning model and an actual trajectory corresponding to the target route data;

updating a parameter of the first route planning model until the first error is smaller than a first threshold; and

determining the updated first route planning model as a target route planning model of the target site, wherein the target route planning model is configured to output a trajectory of a vehicle for autonomous parking in the target site;

wherein the first route planning model is pre-trained based on a set of historical route data, the set of historical route data being associated with a plurality of sites different from the target site and comprising a plurality of first historical route data;

wherein the first route planning model is determined by:

in response to inputting the plurality of first historical route data into a first route planning sub-model, determining a second error between a predicted trajectory outputted by the first route planning sub-model and an actual trajectory corresponding to each first historical route data;

updating a parameter of the first route planning sub-model in a first gradient descent way until the second error is smaller than a second threshold; and

determining the updated first route planning sub-model as the first route planning model;

wherein the method further comprises:

obtaining current route data of the target site by at least one sensor of the vehicle;

determining a trajectory for autonomous parking of the vehicle in the target site by inputting the current route data to the target route planning model; and

controlling steers, throttles, and brakes of the vehicle based on the determined trajectory and parking the vehicle in a predetermined parking space of the target site.

12. The storage medium of claim 11 , wherein the first error is indicated by a target loss function corresponding to the target site, and the second error is indicated by a sum of site loss functions corresponding to the plurality of sites.