Methods and systems for tracking a mover's lane over time
Systems and methods for assigning a lane to an object in an environment of an autonomous vehicle are disclosed. The methods include assigning an instantaneous probability to each of a plurality of lanes in the environment based on a current state of the object, generating a transition matrix for each of the plurality of lanes, and identifying the lane in which the object is moving at the current time t based on the instantaneous probability and the transition matrix. The instantaneous probability is a measure of likelihood that the object is in that lane at a current time. The transition matrix encodes one or more probabilities that the object transitioned either into that lane or out of that lane at the current time.
1. A method of assigning a lane to an object in an environment of an autonomous vehicle, the method comprising:
assigning, based on a current state of the object, an instantaneous probability to each of a plurality of lanes in the environment, the instantaneous probability being a measure of likelihood that the object is in that lane at a current time t;
generating a transition matrix for each of the plurality of lanes, the transition matrix encoding one or more probabilities that the object transitioned either into that lane or out of that lane;
identifying, based on the instantaneous probability and the transition matrix, the lane in which the object is moving at the current time t; and
using the identification of the lane to control navigation of the autonomous vehicle in the environment.
2. The method of claim 1 , wherein identifying the lane in which the object is moving at the current time t comprises identifying one of the plurality of lanes that has a highest assigned probability as the lane in which the object is moving at the current time t, the assigned probability of each lane being a measure of likelihood of the object occupying that lane at the current time step.
3. The method of claim 2 , further comprising determining the assigned probability associated with each of the plurality of lanes by multiplying the instantaneous probability associated with that lane, a previously assigned probability associated with that lane as a measure of likelihood of the object occupying that lane at a previous time step, and a probability of the object transitioning from any of the plurality of lanes into that lane as determined from the transition matrix.
4. The method of claim 1 , wherein generating the transition matrix for each of the plurality of lanes comprises:
receiving a Hidden Markov Model (HMM) for each of the plurality of lanes; and
using the current state of the object and the HMM to generate an initial transition matrix encoding one or more probabilities that the object transitioned to or from that lane from or to another lane in the environment at the current time t.
5. The method of claim 4 , further comprising using relationships between the plurality of lanes for updating the initial transition matrix and generating the transition matrix for each of the plurality of lanes, wherein the relationships between the plurality of lanes include information relating to valid paths for transitioning between each of the plurality of lanes.
6. The method of claim 5 , further comprising determining the relationships between the plurality of lanes using a road network map.
7. The method of claim 4 , further comprising identifying one or more parameters for generating the HMM for each of the plurality of lanes using training data, the training data comprising observed states of a plurality of objects associated with known information relating to lanes occupied by the plurality of objects.
8. The method of claim 1 , wherein assigning the instantaneous probability to each of the plurality of lanes in the environment of the autonomous vehicle comprises determining, using sensor data corresponding to the environment, at least one of the following: percentage of overlap of the object with that lane; alignment of the object with that lane; object classification; direction of travel of the object; speed of the object; acceleration of the object; or pose of the object.
9. The method of claim 1 , wherein assigning the instantaneous probability to each of the plurality of lanes in the environment of the autonomous vehicle comprises using a random forest classifier for assigning the instantaneous probabilities.
10. A system for assigning a lane to an object in an environment of an autonomous vehicle, the system comprising:
at least one processor;
programming instructions stored in a memory and configured to cause the processor to:
assign, based on a current state of the object, an instantaneous probability to each of a plurality of lanes in the environment, the instantaneous probability being a measure of likelihood that the object is in that lane at a current time t;
generate a transition matrix for each of the plurality of lanes, the transition matrix encoding one or more probabilities that the object transitioned either into that lane or out of that lane;
identify, based on the instantaneous probability and the transition matrix, the lane in which the object is moving at the current time t; and
use the identification of the lane to control navigation of the autonomous vehicle in the environment.
11. The system of claim 10 , wherein the programming instructions that cause the processor to identify the lane in which the object is moving at the current time t comprise programming instructions configured to cause the processor to identify one of the plurality of lanes that has a highest assigned probability as the lane in which the object is moving at the current time t, the assigned probability of each lane being a measure of likelihood of the object occupying that lane at the current time step.
12. The system of claim 11 , further comprising programming instructions configured to cause the processor to determine the assigned probability associated with each of the plurality of lanes by multiplying the instantaneous probability associated with that lane, a previously assigned probability associated with that lane as a measure of likelihood of the object occupying that lane at a previous time step, and a probability of the object transitioning from any of the plurality of lanes into that lane as determined from the transition matrix.
13. The system of claim 10 , wherein the programming instructions that cause the processor to generate the transition matrix for each of the plurality of lanes comprise programming instructions configured to cause the processor to:
receive a Hidden Markov Model (HMM) for each of the plurality of lanes; and
use the current state of the object and the HMM to generate an initial transition matrix encoding one or more probabilities that the object transitioned to or from that lane from or to another lane in the environment at the current time t.
14. The system of claim 13 , further comprising programming instructions configured to cause the processor to use relationships between the plurality of lanes for updating the initial transition matrix and generating the transition matrix for each of the plurality of lanes, wherein the relationships between the plurality of lanes include information relating to valid paths for transitioning between each of the plurality of lanes.
15. The system of claim 13 , further comprising programming instructions configured to cause the processor to identify one or more parameters for generating the HMM for each of the plurality of lanes using training data, the training data comprising observed states of a plurality of objects associated with known information relating to lanes occupied by the plurality of objects.
16. The system of claim 10 , wherein the programming instructions that cause the processor to assign the instantaneous probability to each of the plurality of lanes in the environment of the autonomous vehicle comprise programming instructions to cause the processor to determine, using sensor data corresponding to the environment, at least one of the following: percentage of overlap of the object with that lane; alignment of the object with that lane; object classification; direction of travel of the object; speed of the object; acceleration of the object; or pose of the object.
17. The system of claim 10 , wherein the programming instructions that cause the processor to assign the instantaneous probability to each of the plurality of lanes in the environment of the autonomous vehicle comprise programming instructions to cause the processor to use a random forest classifier for assigning the instantaneous probabilities.
18. A computer program product comprising a non-transitory computer-readable medium that stores instructions that, when executed by a computing device, will cause the computing device to perform operations comprising:
assigning, based on a current state of an object, an instantaneous probability to each of a plurality of lanes in an environment of an autonomous vehicle, the instantaneous probability being a measure of likelihood that the object is in that lane at a current time t;
generating a transition matrix for each of the plurality of lanes, the transition matrix encoding one or more probabilities that the object transitioned either into that lane or out of that lane;
identifying, based on the instantaneous probability and the transition matrix, the lane in which the object is moving at the current time t; and
using the identification of the lane to control navigation of the autonomous vehicle in the environment.