Systems and methods for generating motion forecast data for actors with respect to an autonomous vehicle and training a machine learned model for the same

1. A computing system, comprising:

an object detection model configured to receive sensor data, and in response to receipt of the sensor data, generate object detection data describing locations of a plurality of actors relative to an autonomous vehicle;

a graph neural network comprising a plurality of nodes and a plurality of edges, the graph neural network configured to receive the object detection data, and in response to receipt of the object detection data, output motion forecast data with respect to the plurality of actors;

a memory that stores a set of instructions;

one or more processors which use the set of instructions to:

input sensor data into one or more object detection models;

receive, as an output of the one or more object detection models, the object detection data describing the locations of the plurality of the actors relative to the autonomous vehicle;

input the object detection data into the graph neural network;

iteratively update a plurality of node states respectively associated with the plurality of nodes;

receive, as an output of the graph neural network, the motion forecast data with respect to the plurality of actors;

determine a motion plan for the autonomous vehicle based at least in part on the motion forecast data; and

control the autonomous vehicle based at least in part on the motion plan.

2. The computing system of claim 1 , wherein iteratively updating the plurality of nodes states comprises passing a plurality of messages from a transmitting node of the plurality of nodes to a receiving node of the plurality of nodes.

3. The computing system of claim 2 , wherein a number of the plurality of messages that are passed from the transmitting node of the plurality of nodes to the receiving node of the plurality of nodes corresponds with a hyperparameter of the system.

4. The computing system of claim 2 , wherein passing the plurality of messages from the transmitting node of the plurality of nodes to the receiving node of the plurality of nodes comprises transforming the message into a local coordinate system of the respective receiving node.

5. The computing system of claim 1 , wherein iteratively updating the plurality of nodes states comprises, for each respective node of the plurality of nodes, aggregating a plurality of respective messages from each other node of the plurality of nodes.

6. The computing system of claim 5 , wherein the respective messages comprise data describing at least one of relative locations or relative trajectories of the other nodes with respect to the respective node of the plurality of nodes.

7. The computing system of claim 1 , wherein iteratively updating the plurality of the node states comprises, for each respective node of the plurality of node states, updating a respective hidden node state and a respective output node state of the respective node.

8. The computing system of claim 1 , further comprising a plurality of multilayer perceptrons, and wherein iteratively updating the plurality of the node states comprises, for each of the plurality of node states:

inputting the object detection data into the plurality of multilayer perceptrons; and

receiving, as respective outputs of the plurality of multilayer perceptrons, a plurality of output node states.

9. The computing system of claim 1 , wherein the one or more object detection models are further configured to receive map data and generate the object detection data based at least in part on the map data, and wherein the one or more processors further use the set of instructions to input the map data into the one or more object detection models.

10. A computer-implemented method for forecasting actor motion data:

inputting, by a computing system comprising one or more computing devices, sensor data into one or more object detection models configured to receive sensor data, and in response to receipt of the sensor data, generate object detection data describing locations of a plurality of actors relative to an autonomous vehicle;

receiving, by the computing system and as an output of the one or more object detection models, the object detection data describing the location of the plurality of the actors relative to the autonomous vehicle;

inputting, by the computing system, the object detection data into a graph neural network comprising a plurality of nodes and a plurality of edges, the graph neural network configured to receive the object detection data, and in response to receipt of the object detection data, output motion forecast data with respect to the plurality of actors;

iteratively updating, by the computing system, a plurality of node states respectively associated with the plurality of nodes;

receiving, by the computing system and as an output of the graph neural network, the motion forecast data with respect to the plurality of actors;

determine a motion plan for the autonomous vehicle based at least in part on the motion forecast data; and

control the autonomous vehicle based at least in part on the motion plan.

11. The computer-implemented method of claim 10 , wherein iteratively updating the plurality of nodes states comprises passing a plurality of messages from a transmitting node of the plurality of nodes to a receiving node of the plurality of nodes.

12. The computer-implemented method of claim 11 , wherein a number of the plurality of messages that are passed from the transmitting node of the plurality of nodes to the receiving node of the plurality of nodes corresponds with a hyperparameter of the system.

13. The computer-implemented method of claim 11 , wherein passing the plurality of messages from the transmitting node of the plurality of nodes to the receiving node of the plurality of nodes comprises transforming the message into a local coordinate system of the respective receiving node.

14. The computer-implemented method of claim 10 , wherein iteratively updating the plurality of nodes states comprises, for each respective node of the plurality of nodes, aggregating a plurality of respective messages from each other node of the plurality of nodes.

15. The computer-implemented method of claim 14 , wherein the respective messages comprise data describing at least one of relative locations or relative trajectories of the other nodes with respect to the respective node of the plurality of nodes.

16. The computer-implemented method of claim 10 , wherein iteratively updating the plurality of the node states comprises, for each of the plurality of node states, updating a respective hidden node state and a respective output node state.

17. The computer-implemented method of claim 10 , wherein iteratively updating the plurality of the node states comprises, for each of the plurality of node states:

inputting the object detection data into a plurality of multilayer perceptrons; and

receiving, as respective outputs of the plurality of multilayer perceptrons, a plurality of output node states.

18. A computer-implemented method for training a graph neural network for generating actor motion forecast data, the method comprising:

inputting, by a computing system comprising one or more computing devices, sensor data into one or more object detection models configured to receive the sensor data, and in response to receipt of the sensor data, output object detection data describing locations of a plurality of actors relative to an autonomous vehicle;

receiving, by the computing system and as an output of the one or more object detection models, the object detection data describing the location of the plurality of the actors relative to the autonomous vehicle;

inputting, by the computing system, the object detection data into a graph neural network comprising a plurality of nodes and a plurality of edges, the graph neural network configured to receive the object detection data, and in response to receipt of the object detection data, output motion forecast data with respect to the plurality of actors;

iteratively updating, by the computing system, a plurality of node states respectively associated with the plurality of nodes;

receiving, by the computing devices and as an output of the graph neural network, the motion forecast data with respect to the plurality of actors;

adjusting, by the computing system, at least one parameter of the graph neural network based on a comparison of the motion forecast data with respect to ground truth motion forecast data;

determine a motion plan for the autonomous vehicle based at least in part on the motion forecast data; and

control the autonomous vehicle based at least in part on the motion plan.

19. The computer-implemented method of claim 18 , further comprising evaluating, by the computing system, a loss function that describes the comparison of the motion forecast data with respect to ground truth motion forecast data and adjusting, by the computing system, at least one parameter of the one or more object detection models and at least one parameter of the graph neural network based on the evaluation of the loss function.

20. The computer-implemented method of claim 18 , further comprising back-propagating, by the computing system, errors through the graph neural network and the one or more object detection models to determine, by the computing system, a gradient of the loss function to train to the graph neural network and the one or more object detection models in an end-to-end configuration.