Map health monitoring for autonomous systems and applications
In various examples, health of a high definition (HD) map may be monitored to determine whether inaccuracies exist in one or more layers of the HD map. For example, as one or more vehicles rely on the HD map to traverse portions of an environment, disagreements between perception of the one or more vehicles, map layers of the HD map, and/or other disagreement types may be identified and aggregated. Where errors are identified that indicate a drop in health of the HD map, updated data may be crowdsourced from one or more vehicles corresponding to a location of disagreement within the HD map, and the updated data may be used to update, verify, and validate the HD map.
1 . At least one processor comprising:
one or more circuits to:
determine, for a road segment of an HD map, one or more misalignments between a first layer of a version of the HD map and a second layer of the version of the HD map that corresponds to a different class of data than the first layer;
determine a location within the road segment that corresponds to the one or more misalignments;
based at least on the one or more misalignments, transmit, over one or more first network communications to at least one vehicle, at least one indication that triggers the at least one vehicle to generate mapstream data corresponding to the location;
convert, by at least one processor, the mapstream data into at least one map comprising at least one drive segment corresponding to the location;
generate, by the at least one processor and using the at least one drive segment, a fused HD map representation of the drive segment; and
transmit, over one or more second network communications to one or more vehicles, the fused HD map representation, the one or more second network communications causing the one or more vehicles to use the fused HD map representation of the drive segment in an updated version of the HD map to navigate an environment corresponding to the road segment.
2 . The at least one processor of claim 1 , wherein the one or more misalignments are determined based at least on:
determining first localization information corresponding to a first localization performed using the first layer and first perception data;
determining second localization information corresponding to a second localization performed using the second layer and second perception data; and
determining the first localization information disagrees with the second localization information.
3 . The at least one processor of claim 1 , wherein the at least one drive segment includes a plurality of drive segments, the one or more circuits are to geometrically register drive segments of the plurality of drive segments to determine pose links between poses corresponding to drives used to generate the mapstream data, and the generating of the fused HD map representation is based at least on the pose links.
4 . The at least one processor of claim 1 , wherein the at least one drive segment includes a plurality of drive segments, the one or more circuits are to geometrically register drive segments of the plurality of drive segments to determine rotation and translation between frames corresponding to drives used to generate the mapstream data, and the generating of the fused HD map representation is based at least on the rotation and translation between the frames.
5 . The at least one processor of claim 1 , wherein the one or more first network communications trigger a plurality of vehicles to use on-board sensors to collect the mapstream data and upload the mapstream data to at least one server comprising the at least one processor.
6 . The at least one processor of claim 1 , wherein the mapstream data includes at least one of raw sensor data, pre-processed sensor data, perception data generated using one or more deep neural networks (DNNs), or data representative of a relative trajectory.
7 . The at least one processor of claim 1 , wherein the one or more first network communications further immediately invalidate, at the at least one vehicle, one or more portions of the HD map that correspond to the location.
8 . The at least one processor of claim 1 , wherein the fused HD map representation correspond to at least one of the first layer or the second layer.
9 . The at least one processor of claim 1 , wherein the at least one indication triggers execution of a map update workflow including the generation of the mapstream data, generation of the updated version of the HD map, and the one or more vehicles switching to the updated version of the HD map.
10 . The at least one processor of claim 1 , wherein the processor is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation or digital twin operations;
a system for performing collaborative content creation;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system for generating synthetic data;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.
11 . A system comprising:
one or more processing units comprising processing circuitry to:
determine, for a portion of an HD map and based at least on one or more trips of one or more vehicles, one or more misalignments associated with a plurality of map layers of the portion in a version of the HD map;
based at least on the one or more misalignments, transmit, over one or more first network communications to at least one vehicle, at least one indication that triggers the at least one vehicle of the one or more vehicles to generate mapstream data comprising at least one of sensor data, perception data, or relative trajectory data corresponding to the portion of the HD map and the plurality of map layers of the HD map;
convert, by at least one processor, the mapstream data into at least one map comprising at least one drive segment corresponding to the portion;
generate, by the at least one processor and using the at least one drive segment, a fused HD map representation of the drive segment; and
transmit, over one or more second network communications to one or more vehicles, the fused HD map representation, the one or more second network communications causing the one or more vehicles to use the fused HD map representation of the drive segment in an updated version of the HD map to navigate an environment corresponding to the portion of the HD map.
12 . The system of claim 11 , wherein the portion of the HD map corresponds to a road segment, and the plurality of map layers correspond to the road segment.
13 . The system of claim 11 , wherein the processing circuitry is further to transmit the updated portion of the HD map to the one or more vehicles in a flatbuffer format.
14 . The system of claim 11 , wherein the processing circuitry is further to cause the one or more map layers to be deactivated for a first local copy of the HD map on a first vehicle based at least on the one or more misalignments while the one or more maps layers remain active for a second local copy of the HD map on a second vehicle.
15 . The system of claim 11 , wherein the update of the one or more map layers is based at least on one or more first weights indicative of a first safety impact of the one or more misalignments with respect to one or more first map layers of the plurality of map layers and one or more second weights indicative of a second safety impact of the one or more misalignments with respect to one or more second map layers of the plurality of map layers.
16 . The system of claim 11 , wherein the system is comprised in at least one of:
a control system for an autonomous or semi-autonomous machine;
a perception system for an autonomous or semi-autonomous machine;
a system for performing simulation or digital twin operations;
a system for collaborative content creation;
a system for performing deep learning operations;
a system implemented using an edge device;
a system implemented using a robot;
a system for generating synthetic data;
a system incorporating one or more virtual machines (VMs);
a system implemented at least partially in a data center; or
a system implemented at least partially using cloud computing resources.
17 . A method comprising:
based at least on an ego-machine determining one or more misalignments between a plurality of layers of a portion of a local version of an HD map,
generating at least one indication of the one or more misalignments corresponding to the portion of the HD map;
sending, over one or more first network communications, the at least one indication to a remote server to trigger:
the remote server to:
generate, using at least one vehicle, mapstream data corresponding to the portion of the HD map;
convert, by at least one processor, the mapstream data into at least one map comprising at least one drive segment corresponding to the portion;
generate, by the at least one processor and using the at least one drive segment, a fused HD map representation of the drive segment; and
transmit, over one or more second network communications to one or more vehicles, the fused HD map representation, the one or more second network communications causing the one or more vehicles to use the fused HD map representation of the drive segment in an updated version of the HD map to navigate an environment corresponding to the portion of the HD map.
18 . The method of claim 17 , wherein the mapstream data includes at least one of raw sensor data, pre-processed sensor data, perception data generated using one or more deep neural networks (DNNs), or data representative of a relative trajectory.
19 . The method of claim 17 , wherein the portion of the HD map corresponds to a road segment of the HD map, and at least one indication corresponds to the road segment.