Methods and systems for predicting a trajectory of a road agent based on an intermediate space

A computing device comprising:

• a processor;
• and
• a non-transitory computer-readable storage medium comprising instructions that, when executed by the processor, cause the computing device to:
• obtain a respective indicated similarity of trajectories in each of a plurality of combinations of observed vehicle trajectories;
• for each of the combinations,
• calculate a distance between respective intermediate representations, generated by a neural network, of the trajectories in the combination;
• and
• perform a training of the neural network based on a calculated loss between the calculated distance and the indicated similarity of the trajectories in the combination,
• wherein respective calculated losses converge after performing respective trainings for the combinations;
• create an intermediate space of respective intermediate representations, generated by the neural network after performing the respective trainings for the combinations, of observed vehicle trajectories;
• and
• predict a trajectory of a road agent based on a distance between a given intermediate representation in the intermediate space and an intermediate representation of an observed trajectory of the road agent generated by the neural network.

A computing device comprising:

• a processor;
• and
• a non-transitory computer-readable storage medium comprising instructions that, when executed by the processor, cause the computing device to:
• predict a trajectory of a road agent based on a distance between a given intermediate representation in an intermediate space and an intermediate representation of an observed trajectory of the road agent generated by a neural network,
• wherein the intermediate space comprises respective intermediate representations, of observed vehicle trajectories, generated by a neural network after performing respective trainings of the neural network for each of a plurality of combinations of observed vehicle trajectories, and
• wherein the respective trainings comprise, for each of the combinations,
• a calculated a distance between respective intermediate representations, generated by the neural network, for the trajectories in the combination, and
• a training of the neural network based on a calculated loss between the calculated distance and an indicated similarity of the trajectories in the combination, such that respective calculated losses converge after respective trainings for the combinations.