Method and system for determining information on an expected trajectory of an object
A method for determining information on an expected trajectory of an object comprises: determining input data being related to the expected trajectory of the object; determining first intermediate data based on the input data using a machine-learning method; determining second intermediate data based on the input data using a model-based method; and determining the information on the expected trajectory of the object based on the first intermediate data and based on the second intermediate data.
1. A computer-implemented method, the method comprising:
determining, by computer hardware components, information on an expected trajectory of an object by at least:
determining input data related to the expected trajectory of the object;
determining, based on the input data and using a first encoding network, a first latent code corresponding to the expected trajectory of the object;
determining, based on the input data and using a method based on physical laws, results corresponding to the expected trajectory of the object;
determining, based on the results and using a second encoding network, a second latent code corresponding to the expected trajectory of the object;
fusing, using a fusion network, the first latent code and the second latent code to create a fused latent code or a weight;
determining the expected trajectory by:
decoding the fused latent code: or
decoding the first latent code and the second latent code; and
combining the decoded first latent code and the decoded second latent code using the weight; and
operating a vehicle in at least a partially autonomous mode based on the expected trajectory.
2. The computer-implemented method of claim 1 ,
wherein the first encoding network is part of a machine-learned method.
3. The computer-implemented method of claim 1 ,
wherein the input data is based on sensor data, the sensor data comprising at least one of radar data, lidar data, ultrasound data, mobile radio communication data, or camera data.
4. The computer-implemented method of claim 1 ,
wherein the input data comprises information related to positions or velocities or accelerations of the object or further objects.
5. A system, the system comprising:
a computer system comprising a plurality of computer hardware components configured to:
determine information on an expected trajectory of an object by at least:
determining input data related to the expected trajectory of the object;
determining, based on the input data and using a first encoding network, a first latent code corresponding to the expected trajectory of the object;
determining, based on the input data and using a method based on physical laws, results corresponding to the expected trajectory of the object;
determining, based on the results and using a second encoding network, a second latent code corresponding to the expected trajectory of the object;
fusing, using a fusion network, the first latent code and the second latent code to create a fused latent code or weight;
determining the expected trajectory by:
decoding the fused latent code; or
decoding the first latent code and the second latent code; and
combining the decoded first latent code and the decoded second latent code using the weight; and
operating a vehicle in at least a partially autonomous mode based on the expected trajectory.
6. The system of claim 5 , further comprising:
a vehicle comprising the computer system.
7. The system of claim 6 ,
wherein the vehicle further comprises at least one sensor configured to acquire sensor data; and
wherein the computer system is configured to determine the input data based on the acquired sensor data.
8. A non-transitory computer readable medium comprising instructions, that when executed, configure computer hardware components to:
determine information on an expected trajectory of an object by at least:
determining input data related to the expected trajectory of the object;
determining, based on the input data and using a first encoding network, a first latent code corresponding to the expected trajectory of the object;
determining, based on the input data and using a method based on physical laws, results corresponding to the expected trajectory of the object;
determining, based on the results and using a second encoding network, a second latent code corresponding to the expected trajectory of the object;
fusing, using a fusion network, the first latent code and the second latent code to create a fused latent code or a weight;
determining the expected trajectory by:
decoding the fused latent code; or
decoding the first latent code and the second latent code; and
combining the decoded first latent code and the decoded second latent code using the weight; and
operating a vehicle in at least a partially autonomous mode based on the expected trajectory.
9. The computer-implemented method of claim 1 ,
wherein:
the fusing comprises fusing the first latent code and the second latent code to create the fused latent code; and
the determining the expected trajectory comprises decoding the fused latent code.
10. The computer-implemented method of claim 9 ,
wherein the fusion network comprises learnable parameters.
11. The computer-implemented method of claim 1 ,
wherein:
the fusing comprises fusing the first latent code and the second latent code to create the weight; and
the determining the expected trajectory comprises:
decoding the first latent code and the second latent code; and
combining the decoded first latent code and the decoded second latent code using the weight.
12. The computer-implemented method of claim 11 ,
wherein the combining comprises a weighted sum.
13. The computer-implemented method of claim 11 ,
wherein:
the decoded first latent code comprises a first portion of the expected trajectory; and
the decoded second latent code comprises a second portion of the expected trajectory.
14. The system of claim 5 ,
wherein:
the fusing comprises fusing the first latent code and the second latent code to create the fused latent code; and
the determining the expected trajectory comprises decoding the fused latent code.
15. The system of claim 14 ,
wherein the fusion network comprises learnable parameters.
16. The system of claim 5 ,
wherein:
the fusing comprises fusing the first latent code and the second latent code to create the weight; and
the determining the expected trajectory comprises:
decoding the first latent code and the second latent code; and
combining the decoded first latent code and the decoded second latent code using the weight.
17. The system of claim 16 ,
wherein the combining comprises a weighted sum.
18. The system of claim 16 ,
wherein:
the decoded first latent code comprises a first portion of the expected trajectory; and
the decoded second latent code comprises a second portion of the expected trajectory.
19. The non-transitory computer readable medium of claim 8 ,
wherein:
the fusing comprises fusing the first latent code and the second latent code to create the fused latent code; and
the determining the expected trajectory comprises decoding the fused latent code.
20. The non-transitory computer readable medium of claim 8 ,
wherein:
the fusing comprises fusing the first latent code and the second latent code to create the weight; and
the determining the expected trajectory comprises:
decoding the first latent code and the second latent code; and
combining the decoded first latent code and the decoded second latent code using the weight.