Latent variable determination by a diffusion model
A computing device can implement techniques for predicting an object trajectory or scene information. For example, the techniques may include inputting latent variable data into a machine learned model. The machine learned model may output an object trajectory (e.g., position data, velocity data, acceleration data, etc.) for one or more objects in the environment based on the latent variable data. The object trajectory can be sent to a vehicle computing device for consideration during vehicle planning, which may include simulation.
1. A method comprising:
receiving, by a diffusion model, map data representing an environment, wherein the diffusion model comprises one or more self-attention layers;
receiving, by the diffusion model, condition data representing a state or an action of an object in the environment;
determining, by the diffusion model and based at least in part on the map data, the condition data, and output data associated with the one or more self-attention layers, latent variable data associated with the object;
inputting the latent variable data into a machine learned model;
determining, by the machine learned model and based at least in part on the latent variable data, an object trajectory for the object to follow in the environment; and
causing control of a vehicle in the environment by executing one or more vehicle maneuvers based at least in part on the object trajectory to avoid potential interaction between the vehicle and the object.
2. The method of claim 1 , wherein:
the machine learned model comprises at least one of a Generative Adversarial Network (GAN), a Graph Neural Network (GNN), a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), or a transformer model.
3. The method of claim 1 , wherein:
the diffusion model is trained based at least in part on a set of conditions, at least one condition of the set of conditions comprising a previous action, a previous position, a previous trajectory, or a previous acceleration of the object.
4. The method of claim 1 , wherein:
the condition data comprises a token from a codebook associated with a transformer model, and
the token representing a behavior of the object.
5. The method of claim 1 , wherein:
the condition data comprises a node from a Graph Neural Network, and
the node representing a behavior of the object.
6. The method of claim 1 , wherein:
the condition data represents one of: a yield action, a drive straight action, a left turn action, a right turn action, a brake action, an acceleration action, a steering action, a lane change action, a position, a heading, or an acceleration of the object.
7. The method of claim 1 , wherein:
the diffusion model is trained by incrementally denoising data to generate an output based on a conditional input.
8. The method of claim 1 , further comprising:
mapping the latent variable data to continuous variables associated with feature vectors,
wherein inputting the latent variable data into the machine learned model comprises inputting the continuous variables associated with the feature vectors, and
determining, by the machine learned model, the object trajectory is based at least in part on the continuous variables associated with the feature vectors.
9. The method of claim 1 , wherein:
the object is a first object,
the condition data represents a potential action of a second object in the environment, and
the latent variable data represents a potential interaction between the second object and the first object.
10. The method of claim 1 , wherein:
the environment is a simulated environment,
the condition data represents a feature of the simulated environment, and further comprising:
determining, by the machine learned model and based at least in part on the feature of the simulated environment, scene data for testing or verifying a scenario in the simulated environment.
11. One or more non-transitory computer-readable media storing instructions that, when executed, cause one or more processors to perform actions comprising:
receiving, by a diffusion model, map data representing an environment, wherein the diffusion model comprises one or more self-attention layers;
receiving, by the diffusion model, condition data representing a state or an action of an object in the environment;
determining, by the diffusion model and based at least in part on the map data, the condition data, and output data associated with the one or more self-attention layers, latent variable data associated with the object;
inputting the latent variable data into a machine learned model;
determining, by the machine learned model and based at least in part on the latent variable data, an object trajectory for the object to follow in the environment; and
causing control of a vehicle in the environment by executing one or more vehicle maneuvers based at least in part on the object trajectory to avoid potential interaction between the vehicle and the object.
12. The one or more non-transitory computer-readable media of claim 11 , wherein:
the machine learned model comprises at least one of a Generative Adversarial Network (GAN), a Graph Neural Network (GNN), a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), or a transformer model.
13. The one or more non-transitory computer-readable media of claim 11 , wherein:
the diffusion model is trained based at least in part on a set of conditions, at least one condition of the set of conditions comprising a previous action, a previous position, a previous trajectory, or a previous acceleration of the object.
14. The one or more non-transitory computer-readable media of claim 11 ,
wherein:
the object is a first object,
the condition data represents a potential action of a second object in the environment, and
the latent variable data represents a potential interaction between the second object and the first object.
15. A system comprising:
one or more processors; and
one or more non-transitory computer-readable media storing instructions executable by the one or more processors, wherein the instructions, when executed, cause the one or more processors to perform actions comprising:
receiving, by a diffusion model, map data representing an environment, wherein the diffusion model comprises one or more self-attention layers;
receiving, by the diffusion model, condition data representing a state or an action of an object in the environment;
determining, by the diffusion model and based at least in part on the map data, the condition data, and output data associated with the one or more self-attention layers, latent variable data associated with the object;
inputting the latent variable data into a machine learned model;
determining, by the machine learned model and based at least in part on the latent variable data, an object trajectory for the object to follow in the environment; and
causing control of a vehicle in the environment by executing one or more vehicle maneuvers based at least in part on the object trajectory to avoid potential interaction between the vehicle and the object.
16. The system of claim 15 , wherein:
the machine learned model comprises at least one of a Generative Adversarial Network (GAN), a Graph Neural Network (GNN), a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), or a transformer model.
17. The system of claim 15 , wherein:
the diffusion model is trained based at least in part on a set of conditions, at least one condition of the set of conditions comprising a previous action, a previous position, a previous trajectory, or a previous acceleration of the object.
18. The system of claim 15 , wherein:
the condition data comprises a token from a codebook associated with a transformer model, and
the token representing a behavior of the object.
19. The system of claim 15 , wherein:
the condition data comprises a node from a Graph Neural Network, and
the node representing a behavior of the object.
20. The system of claim 15 , wherein:
the condition data represents one of: a yield action, a drive straight action, a left turn action, a right turn action, a brake action, an acceleration action, a steering action, a lane change action, a position, a heading, or an acceleration of the object.