Patent Application
20250077856
The present disclosure generally relates to subsampling data for model training, and more specifically to subsampling data based on error detection associated with executing a model.
An autonomous vehicle is a motorized vehicle that can navigate without a human driver. An exemplary autonomous vehicle can include various sensors, such as a camera sensor, a light detection and ranging (LIDAR) sensor, and a radio detection and ranging (RADAR) sensor, amongst others. The sensors collect data and measurements that the autonomous vehicle can use for operations such as navigation. The sensors can provide the data and measurements to an internal computing system of the autonomous vehicle, which can use the data and measurements to control a mechanical system of the autonomous vehicle, such as a vehicle propulsion system, a braking system, or a steering system. Typically, the sensors are mounted at fixed locations on the autonomous vehicles.
The various advantages and features of the present technology will become apparent by reference to specific implementations illustrated in the appended drawings. A person of ordinary skill in the art will understand that these drawings only show some examples of the present technology and would not limit the scope of the present technology to these examples. Furthermore, the skilled artisan will appreciate the principles of the present technology as described and explained with additional specificity and detail through the use of the accompanying drawings in which:
The detailed description set forth below is intended as a description of various configurations of the subject technology and is not intended to represent the only configurations in which the subject technology can be practiced. The appended drawings are incorporated herein and constitute a part of the detailed description. The detailed description includes specific details for the purpose of providing a more thorough understanding of the subject technology. However, it will be clear and apparent that the subject technology is not limited to the specific details set forth herein and may be practiced without these details. In some instances, structures and components are shown in block diagram form to avoid obscuring the concepts of the subject technology.
Some aspect of the present technology may relate to the gathering and use of data available from various sources to improve safety, quality, and experience. The present disclosure contemplates that in some instances, this gathered data may include personal information. The present disclosure contemplates that the entities involved with such personal information respect and value privacy policies and practices.
Autonomous vehicles (AVs) can navigate roadways without a human driver based on sensor signals generated by multiple sensor systems, as described in more detail below. The sensor systems can include one or more types of sensors that can be arranged about the AV, including but not limited to camera sensors. In some examples, the AV can interpret sensor signals to detect and classify objects in the environment using a perception stack, as explained in more detail below. The perception stack can enable the AV to “see” (e.g., via cameras, LIDAR sensors, infrared sensors, etc.), “hear” (e.g., via microphones, ultrasonic sensors, RADAR, etc.), and “feel” (e.g., pressure sensors, force sensors, impact sensors, etc.) its environment using information from the sensor systems as well as other data sources. In some examples, the AV controller relies on the correct detection and classification of the objects in the AVs environment to subsequently provide commands for the actuators that control the AV's steering, throttle, brake, and drive unit. It is therefore crucial that the AV's perception stack can correctly detect and classify objects in the AV's environment so that the AV planner and controller can determine and implement a plan for controlling the AV in a safe and effective manner.
AVs can be controlled through software stacks that implement machine learning techniques to control the AVs based on sensor data that is captured during operation of the AVs. Specifically, software stacks can interpret gathered sensor data to perceive an environment and formulate a plan for controlling the AV based on perceived objects in the environment. In turn, the AV can be controlled according to this plan to facilitate operation of the AV in the environment.
Software stacks that implement machine learning techniques to control AVs can be trained through large amounts of data, e.g. labeled data. However, this is problematic as training models on such large datasets consumes large amounts of computational resources over great amounts of time. Further, as models are improved and trained to operate under different scenarios, the models need to be trained on increasing amounts of data. As follows, this further increases the amount of time and computational resources that are consumed in training such models.
With respect to AVs, data is captured by various sensors at set rates, e.g. 1 frame per second. The resulting sequence of frames are referred to as segments. However, such segments contain more data than is actually needed to train models, e.g. for controlling the AVs.
In order to decrease the training times of models and resources used in training the models, the amount of data that is used to actually train the models can be reduced by sampling such data and then training the models with the samples. However, datasets for model training purposes are often sampled in a non-optimal manner. For example, datasets can be uniformly subsampled, e.g. 5x, which may not be enough to accurately train the model. In another example, long-tail datasets are not subsampled and instead used in the entirety to train a model, when only a few scenes per segment are useful for training the model. In yet another example, datasets can be oversampled in an attempt to oversample a few scenes of interest per segment.
The disclosed technology addresses the problems associated with training models with large datasets. Specifically and as will be discussed in greater detail later, data can be subsampled for training a model based on error detection associated with execution of the model.
The perception process 102 functions to access sensor data gathered by an AV. The perception process 102 can fuse the sensor data. From the sensor data, the perception process 102 can track objects. Specifically, the perception process 102 can identify where tracked objects are in a field of view, e.g. relative to the AV.
The prediction process 104 functions to predict where objects will be in a field of view. Specifically, the prediction process 104 can predict the location of objects that are not tracked by the perception process 102. The prediction process 104 can predict the location of objects based on the tracked object output of the perception process 102.
The planner process 106 functions to identify a path for the AV. Specifically, the planner process 106 functions to identify a path for the AV based on either or both the output of the perception process 102 and the prediction process 104. In identifying a path for the AV, the planner process can weigh various moves by the AV against costs with respect to the output of either or both the perception process 102 and the prediction process 104.
The motion planner process 108 functions to identify a refined path for the AV. In particular, the motion planner process 108 functions to identify a refined path for the AV with respect to the path identified by the planner process 106. A refined path developed by the motion planner process 108 can include a path that is planned according to smaller time operations and smaller distances in comparison to the scheme that is used to develop the path by the planner process 106.
The control process 110 functions to communicate with control systems of the AV to implement the plan developed by either or both the planner process 106 and the motion planner process 108. Specifically, the control process 110 can communicate values of parameters for controlling the AV to applicable systems for controlling the AV. For example, the control process 110 can specify to an acceleration controller of the AV to accelerate at 10%.
The disclosure now continues with a discussion of subsampling data for model training based on error detection associated with executing a model. Specifically,
The data for training the model 202 can be split into the first subset of data 204 and the second subset of data 206 through one or more applicable techniques. Specifically, the data for training the model 202 can be split into the first subset of data 204 and the second subset of data 206 based on temporal factors. For example, the first 20 minutes in a captured on-road scene can be separated from the last 20 minutes of the captured on-road scene. Further, the data for training the model 202 can be split according to the different contexts associated with the data for training the model 202. Contexts associated with data, as used herein, can include applicable factors related to the generation of the data and the semantic meaning of the data. For example, if the data for training the model 202 was gathered in different environments, then the data for training the model 202 can be split into the first subset of data 204 and the second subset of data 206 according to the different environments in which the data for training the model 202 was generated.
The data for training the model 202 can be filtered from a larger dataset before being split into the first subset of data 204 and the second subset of data 206. Specifically, the data for training the model 202 can be filtered from a larger dataset based on contexts associated with either the data for training the model 202 and the larger dataset from which it is filtered. For example, the data for training the model 202 can be filtered from a larger dataset to include scenes in which a false positive object recognition occurs within 3 meters of an AV. In this example, the data for training the model 202 can correspond to a liquid being present on one or more sensors of the AV that is falsely detected as an object, otherwise a false positive. In another example, the data for training the model 202 can be filtered from a larger dataset to include scenes in which a false negative object recognition occurs within 20 meters of an AV. In this example, the data for training the model 202 can correspond to fog in the vicinity of the AV and a failure to detect the object, otherwise a false negative, because of the fog. In another example, the data for training the model 202 can be filtered from a larger dataset to include scenes with high amounts of two wheel vehicles and/or pedestrians. In yet another example, the data for training the model 202 can be filtered from a larger dataset to include detected objects, both falsely detected and correctly detected, in the ultra nearfield of an AV.
After the data for training the model 202 is separated into the first subset of data 204 and the second subset of data 206, then the first subset of data 204 is used to train the model and generate a trained model 208. Specifically, the first subset of data 204 can be labeled and used to train the model. The first subset of data 204 can be labeled using an applicable process. For example, human labelers can inspect AV road data of the first subset of data 204 to identify scenes where a false positive object detection, detection of an object when one does not actually exist, occurs due to liquid on a sensor of an AV. In turn, the labeled data of liquid covering the sensors to cause false positives can be input for training a perception model and generating a trained perception model.
With respect to the second subset of data 206, portions of the second subset of data 206 can be mined and used to train or otherwise analyze performance of the trained model 208. This can be done with only portions of the second subset of data 206 while refraining from using the entirety of the second subset of data 206. By using only portions of the second subset of data 206 to train and analyze the trained model 208, computational costs can be reduced and time can be saved in comparison to training and analyzing the trained model 208 on larger sets of data.
In using only a portion of the second subset of data 206 to train the model, the second subset of data 206 can be sampled, e.g. as part of subsampling the data for training the model 202. Specifically, the second subset of data 206 can be sampled based on replay analysis applied to the second subset of data 206. More specifically, the second subset of data 206 can be sampled based on replay analysis associated with running the trained model 208 on the second subset of data 206, e.g. as part of identifying one or more errors associated with running the trained model on the second subset of data 206. Further, the second subset of data 206 can be sampled for training the model 202 through another applicable analysis technique that is distinct from replay analysis, e.g. based on the original data logs themselves that make up the second subset of data 206.
Replay analysis associated with running the trained model 208 on the second subset of data 206 can include applicable analysis that is performed without actually running the trained model 208 on the entirety of the second subset of data 206. Specifically, replay analysis associated with running the trained model 208 can include analysis that is performed to identify portions of the second subset of data 206 that include one or more error modes that could be present if the trained model were run on the entire second subset of data. The errors do not necessarily need to actually be present in the identified portions of the data, just potentially be present in the identified portions of the data, e.g. as an error mode. For example, replay analysis could include identifying a scene in the second subset of data 206 in which a false positive detection of an object is caused by rain on a sensor of an AV. Further in the example, the replay analysis could also include identifying a scene in the second subset of data 206 in which an object is correctly detected, even though there is rain on the sensor of the AV, otherwise the error mode.
Replay analysis associated with running the trained model 208 on the second subset of data 206, e.g. as part of identifying one or more errors associated with running the trained model on the second subset of data 206, can be performed by an error miner. An error miner can function to identify error modes for errors by leveraging replay analysis information that exists for the errors. Specifically, an error miner can associate specific characteristics of errors that are present in captured AV data, as identified through replay analysis, with the errors themselves. Such characteristics define the error modes for the corresponding errors which are ultimately used to find the errors themselves in sampled data. Example characteristics of errors can include, false positive detection of a car within 3 meters of an AV, a misclassification of a bike as a human, multiple ML models missing a particular object, and missing detection for object labeled with special tags, e.g. critical objects and flying objects.
Error miners can be specific to an error type. In particular, an error miner can be configured to detect a specific type of error. For example, an error miner can be configured to detect false negative detections caused by fog in an AV environment. In turn, a variety of error miners can be applied to data to identify error modes for different types of errors in the data. In being configured to detect a specific type of error, an error miner can include a filter that is applied to filter data from a larger dataset that presents an error mode of a specific type of error. For example, an error miner for identifying scenes where an object is not detected in an AV environment because of fog can include a filter that separates such scenes from a larger dataset of scenes in the AV environment.
Error miners for detecting errors in the trained model can be generated, or otherwise trained, based on execution of the trained model, e.g. through replay analysis of execution of the trained model. Specifically, the trained model 208 can be applied to data, as part of model inference, and the output of such inference can be replayed to identify characteristics associated with errors. In turn, the characteristics associated with the errors can be used to train error miners for such errors.
The error miners can be used to perform error mining on the second subset of data 206 and extract a second subset of data that presented the error mode 210 from the second subset of data 206. The second subset of data that presents the error mode 210 can then be used in further refining the trained model 208. Specifically, the trained model 208 can be run on the second subset of data that presents the error mode 210 as part of model inference. In running the trained model 208 on the second subset of data that presents the error mode 210, the portions of this data that actually present the one or more errors associated with the error mode can be identified. For example, if an error is falsely detecting an object when there is liquid on a sensor of an AV, then the model can be run on the data to separate the instances where an object is falsely detected and an object is correctly not detected when there is liquid on a sensor of the AV.
The identified portions of the data that actually present the one or more errors associated with the error mode can be used in further refining the trained model 208. Specifically, the data that actually present the one or more errors can be used in further refining the trained model 208 to generate a refined trained model 212. In further refining the trained model 208 to generate the refined trained model 212, the model can be trained based on the data that actually present the one or more errors. Specifically, the data that actually presents the one or more errors can be labeled and used in further training the trained model 208. Additionally, the model can be trained based on the data that is identified as presenting the error mode without actually including the one or more errors. Specifically, the data that presents the error mode without presenting the one or more errors can be labeled and used in further training the trained model 208.
Further, in refining the trained model 208 based on the second subset of data that presents the error mode 210, the second subset of data that presents the error mode 210 can be used to update replay test results associated with the trained model 208. In turn, such updated replay test results can be used in further refining the trained model 208 in later instances.
The disclosure now continues with a further discussion of subsampling data for training a model based on error mode detection. Specifically,
At operation 302, data for training a model is accessed. The data for training the model can include real-world road data that is gathered by an AV operating in a real-world environment. The data can be labeled or unlabeled. Examples of AV data for training a model can include large vehicle data and splash data corresponding to liquid being present on sensors of an AV.
At operation 304, the data is separated into a first subset of data and a second subset of data. The data can be separated into the first subset of data and the second subset of data using an applicable technique, such as the techniques described herein. Specifically, the data can be separated based on temporal factors associated with the data.
At operation 306, the model is trained with the first subset of data to generate a trained model. Specifically, the model can be trained while refraining from training the model on the second subset of data in its entirety. This can conserve computational resources that would be used in training the model on the second subset of data in its entirety.
At operation 308, one or more errors associated with running the trained model on the second subset of data are identified. Errors can include applicable errors that can present when running the trained model on data, such as the errors described herein. For example, the one or more errors can include the model failing to produce a bounding box when an object exists in scene data gathered by the AV. In another example, the one or more errors can include the model producing a bounding box when an object does not exist in scene data gathered by the AV.
At operation 310, the second subset of data is filtered to generate filtered data associated with the one or more errors. Specifically, a specific type of error that is capable of occurring in applying the model can be selected. Then, the second subset of data is mined to identify portions of the data that present the error mode corresponding to the specific type of error that is selected.
At operation 312, the trained model is further trained based on the filtered data to generate a refined trained model. Specifically, the model can be trained based on instances in the filtered data that actually present the error. Further, the model can be trained based on instances in the filtered data that present the error mode but do not lead to the error occurring.
The disclosure now turns to a discussion of an AV environment. Specifically,
In this example, the AV environment 400 includes an AV 402, a data center 450, and a client computing device 470. The AV 402, the data center 450, and the client computing device 470 can communicate with one another over one or more networks (not shown), such as a public network (e.g., the Internet, an Infrastructure as a Service (IaaS) network, a Platform as a Service (PaaS) network, a Software as a Service (SaaS) network, other Cloud Service Provider (CSP) network, etc.), a private network (e.g., a Local Area Network (LAN), a private cloud, a Virtual Private Network (VPN), etc.), and/or a hybrid network (e.g., a multi-cloud or hybrid cloud network, etc.).
The AV 402 can navigate roadways without a human driver based on sensor signals generated by multiple sensor systems 404, 406, and 408. The sensor systems 404-408 can include one or more types of sensors and can be arranged about the AV 402. For instance, the sensor systems 404-408 can include Inertial Measurement Units (IMUs), cameras (e.g., still image cameras, video cameras, etc.), light sensors (e.g., LIDAR systems, ambient light sensors, infrared sensors, etc.), RADAR systems, GPS receivers, audio sensors (e.g., microphones, Sound Navigation and Ranging (SONAR) systems, ultrasonic sensors, etc.), engine sensors, speedometers, tachometers, odometers, altimeters, tilt sensors, impact sensors, airbag sensors, seat occupancy sensors, open/closed door sensors, tire pressure sensors, rain sensors, and so forth. For example, the sensor system 404 can be a camera system, the sensor system 406 can be a LIDAR system, and the sensor system 408 can be a RADAR system. Other examples may include any other number and type of sensors.
The AV 402 can also include several mechanical systems that can be used to maneuver or operate the AV 402. For instance, the mechanical systems can include a vehicle propulsion system 430, a braking system 432, a steering system 434, a safety system 436, and a cabin system 438, among other systems. The vehicle propulsion system 430 can include an electric motor, an internal combustion engine, or both. The braking system 432 can include an engine brake, brake pads, actuators, and/or any other suitable componentry configured to assist in decelerating the AV 402. The steering system 434 can include suitable componentry configured to control the direction of movement of the AV 402 during navigation. The safety system 436 can include lights and signal indicators, a parking brake, airbags, and so forth. The cabin system 438 can include cabin temperature control systems, in-cabin entertainment systems, and so forth. In some examples, the AV 402 might not include human driver actuators (e.g., steering wheel, handbrake, foot brake pedal, foot accelerator pedal, turn signal lever, window wipers, etc.) for controlling the AV 402. Instead, the cabin system 438 can include one or more client interfaces (e.g., Graphical User Interfaces (GUIs), Voice User Interfaces (VUIs), etc.) for controlling certain aspects of the mechanical systems 430-438.
The AV 402 can include a local computing device 410 that is in communication with the sensor systems 404-408, the mechanical systems 430-438, the data center 450, and the client computing device 470, among other systems. The local computing device 410 can include one or more processors and memory, including instructions that can be executed by the one or more processors. The instructions can make up one or more software stacks or components responsible for controlling the AV 402; communicating with the data center 450, the client computing device 470, and other systems; receiving inputs from riders, passengers, and other entities within the AV's environment; logging metrics collected by the sensor systems 404-408; and so forth. In this example, the local computing device 410 includes a perception stack 412, a localization stack 414, a prediction stack 416, a planning stack 418, a communications stack 420, a control stack 422, an AV operational database 424, and an HD geospatial database 426, among other stacks and systems.
Perception stack 412 can enable the AV 402 to “see” (e.g., via cameras, LIDAR sensors, infrared sensors, etc.), “hear” (e.g., via microphones, ultrasonic sensors, RADAR, etc.), and “feel” (e.g., pressure sensors, force sensors, impact sensors, etc.) its environment using information from the sensor systems 404-408, the localization stack 414, the HD geospatial database 426, other components of the AV, and other data sources (e.g., the data center 450, the client computing device 470, third party data sources, etc.). The perception stack 412 can detect and classify objects and determine their current locations, speeds, directions, and the like. In addition, the perception stack 412 can determine the free space around the AV 402 (e.g., to maintain a safe distance from other objects, change lanes, park the AV, etc.). The perception stack 412 can identify environmental uncertainties, such as where to look for moving objects, flag areas that may be obscured or blocked from view, and so forth. In some examples, an output of the perception stack 412 can be a bounding area around a perceived object that can be associated with a semantic label that identifies the type of object that is within the bounding area, the kinematic of the object (information about its movement), a tracked path of the object, and a description of the pose of the object (its orientation or heading, etc.).
Localization stack 414 can determine the AV's position and orientation (pose) using different methods from multiple systems (e.g., GPS, IMUs, cameras, LIDAR, RADAR, ultrasonic sensors, the HD geospatial database 426, etc.). For example, in some cases, the AV 402 can compare sensor data captured in real-time by the sensor systems 404-408 to data in the HD geospatial database 426 to determine its precise (e.g., accurate to the order of a few centimeters or less) position and orientation. The AV 402 can focus its search based on sensor data from one or more first sensor systems (e.g., GPS) by matching sensor data from one or more second sensor systems (e.g., LIDAR). If the mapping and localization information from one system is unavailable, the AV 402 can use mapping and localization information from a redundant system and/or from remote data sources.
Prediction stack 416 can receive information from the localization stack 414 and objects identified by the perception stack 412 and predict a future path for the objects. In some examples, the prediction stack 416 can output several likely paths that an object is predicted to take along with a probability associated with each path. For each predicted path, the prediction stack 416 can also output a range of points along the path corresponding to a predicted location of the object along the path at future time intervals along with an expected error value for each of the points that indicates a probabilistic deviation from that point.
Planning stack 418 can determine how to maneuver or operate the AV 402 safely and efficiently in its environment. For example, the planning stack 418 can receive the location, speed, and direction of the AV 402, geospatial data, data regarding objects sharing the road with the AV 402 (e.g., pedestrians, bicycles, vehicles, ambulances, buses, cable cars, trains, traffic lights, lanes, road markings, etc.) or certain events occurring during a trip (e.g., emergency vehicle blaring a siren, intersections, occluded areas, street closures for construction or street repairs, double-parked cars, etc.), traffic rules and other safety standards or practices for the road, user input, and other relevant data for directing the AV 402 from one point to another and outputs from the perception stack 412, localization stack 414, and prediction stack 416. The planning stack 418 can determine multiple sets of one or more mechanical operations that the AV 402 can perform (e.g., go straight at a specified rate of acceleration, including maintaining the same speed or decelerating; turn on the left blinker, decelerate if the AV is above a threshold range for turning, and turn left; turn on the right blinker, accelerate if the AV is stopped or below the threshold range for turning, and turn right; decelerate until completely stopped and reverse; etc.), and select the best one to meet changing road conditions and events. If something unexpected happens, the planning stack 418 can select from multiple backup plans to carry out. For example, while preparing to change lanes to turn right at an intersection, another vehicle may aggressively cut into the destination lane, making the lane change unsafe. The planning stack 418 could have already determined an alternative plan for such an event. Upon its occurrence, it could help direct the AV 402 to go around the block instead of blocking a current lane while waiting for an opening to change lanes.
Control stack 422 can manage the operation of the vehicle propulsion system 430, the braking system 432, the steering system 434, the safety system 436, and the cabin system 438. The control stack 422 can receive sensor signals from the sensor systems 404-408 as well as communicate with other stacks or components of the local computing device 410 or a remote system (e.g., the data center 450) to effectuate operation of the AV 402. For example, the control stack 422 can implement the final path or actions from the multiple paths or actions provided by the planning stack 418. This can involve turning the routes and decisions from the planning stack 418 into commands for the actuators that control the AV's steering, throttle, brake, and drive unit.
Communications stack 420 can transmit and receive signals between the various stacks and other components of the AV 402 and between the AV 402, the data center 450, the client computing device 470, and other remote systems. The communications stack 420 can enable the local computing device 410 to exchange information remotely over a network, such as through an antenna array or interface that can provide a metropolitan WIFI network connection, a mobile or cellular network connection (e.g., Third Generation (3G), Fourth Generation (4G), Long-Term Evolution (LTE), 5th Generation (5G), etc.), and/or other wireless network connection (e.g., License Assisted Access (LAA), Citizens Broadband Radio Service (CBRS), MULTEFIRE, etc.). Communications stack 420 can also facilitate the local exchange of information, such as through a wired connection (e.g., a user's mobile computing device docked in an in-car docking station or connected via Universal Serial Bus (USB), etc.) or a local wireless connection (e.g., Wireless Local Area Network (WLAN), Low Power Wide Area Network (LPWAN), Bluetooth®, infrared, etc.).
The HD geospatial database 426 can store HD maps and related data of the streets upon which the AV 402 travels. In some examples, the HD maps and related data can comprise multiple layers, such as an areas layer, a lanes and boundaries layer, an intersections layer, a traffic controls layer, and so forth. The areas layer can include geospatial information indicating geographic areas that are drivable (e.g., roads, parking areas, shoulders, etc.) or not drivable (e.g., medians, sidewalks, buildings, etc.), drivable areas that constitute links or connections (e.g., drivable areas that form the same road) versus intersections (e.g., drivable areas where two or more roads intersect), and so on. The lanes and boundaries layer can include geospatial information of road lanes (e.g., lane centerline, lane boundaries, type of lane boundaries, etc.) and related attributes (e.g., direction of travel, speed limit, lane type, etc.). The lanes and boundaries layer can also include three-dimensional (3D) attributes related to lanes (e.g., slope, elevation, curvature, etc.). The intersections layer can include geospatial information of intersections (e.g., crosswalks, stop lines, turning lane centerlines and/or boundaries, etc.) and related attributes (e.g., permissive, protected/permissive, or protected only left turn lanes; legal or illegal u-turn lanes; permissive or protected only right turn lanes; etc.). The traffic controls lane can include geospatial information of traffic signal lights, traffic signs, and other road objects and related attributes.
AV operational database 424 can store raw AV data generated by the sensor systems 404-408, stacks 412-422, and other components of the AV 402 and/or data received by the AV 402 from remote systems (e.g., the data center 450, the client computing device 470, etc.). In some examples, the raw AV data can include HD LIDAR point cloud data, image data, RADAR data, GPS data, and other sensor data that the data center 450 can use for creating or updating AV geospatial data or for creating simulations of situations encountered by AV 402 for future testing or training of various machine learning algorithms that are incorporated in the local computing device 410.
Data center 450 can include a private cloud (e.g., an enterprise network, a co-location provider network, etc.), a public cloud (e.g., an Infrastructure as a Service (IaaS) network, a Platform as a Service (PaaS) network, a Software as a Service (SaaS) network, or other Cloud Service Provider (CSP) network), a hybrid cloud, a multi-cloud, and/or any other network. The data center 450 can include one or more computing devices remote to the local computing device 410 for managing a fleet of AVs and AV-related services. For example, in addition to managing the AV 402, the data center 450 may also support a ride-hailing service (e.g., a ridesharing service), a delivery service, a remote/roadside assistance service, street services (e.g., street mapping, street patrol, street cleaning, street metering, parking reservation, etc.), and the like.
Data center 450 can send and receive various signals to and from the AV 402 and the client computing device 470. These signals can include sensor data captured by the sensor systems 404-408, roadside assistance requests, software updates, ride-hailing/ridesharing pick-up and drop-off instructions, and so forth. In this example, the data center 450 includes a data management platform 452, an Artificial Intelligence/Machine Learning (AI/ML) platform 454, a simulation platform 456, a remote assistance platform 458, and a ride-hailing platform 460, and a map management platform 462, among other systems.
Data management platform 452 can be a “big data” system capable of receiving and transmitting data at high velocities (e.g., near real-time or real-time), processing a large variety of data and storing large volumes of data (e.g., terabytes, petabytes, or more of data). The varieties of data can include data having different structures (e.g., structured, semi-structured, unstructured, etc.), data of different types (e.g., sensor data, mechanical system data, ride-hailing service, map data, audio, video, etc.), data associated with different types of data stores (e.g., relational databases, key-value stores, document databases, graph databases, column-family databases, data analytic stores, search engine databases, time series databases, object stores, file systems, etc.), data originating from different sources (e.g., AVs, enterprise systems, social networks, etc.), data having different rates of change (e.g., batch, streaming, etc.), and/or data having other characteristics. The various platforms and systems of the data center 450 can access data stored by the data management platform 452 to provide their respective services.
The AI/ML platform 454 can provide the infrastructure for training and evaluating machine learning algorithms for operating the AV 402, the simulation platform 456, the remote assistance platform 458, the ride-hailing platform 460, the map management platform 462, and other platforms and systems. Using the AI/ML platform 454, data scientists can prepare data sets from the data management platform 452; select, design, and train machine learning models; evaluate, refine, and deploy the models; maintain, monitor, and retrain the models; and so on.
Simulation platform 456 can enable testing and validation of the algorithms, machine learning models, neural networks, and other development efforts for the AV 402, the remote assistance platform 458, the ride-hailing platform 460, the map management platform 462, and other platforms and systems. Simulation platform 456 can replicate a variety of driving environments and/or reproduce real-world scenarios from data captured by the AV 402, including rendering geospatial information and road infrastructure (e.g., streets, lanes, crosswalks, traffic lights, stop signs, etc.) obtained from a cartography platform (e.g., map management platform 462); modeling the behavior of other vehicles, bicycles, pedestrians, and other dynamic elements; simulating inclement weather conditions, different traffic scenarios; and so on.
Remote assistance platform 458 can generate and transmit instructions regarding the operation of the AV 402. For example, in response to an output of the AI/ML platform 454 or other system of the data center 450, the remote assistance platform 458 can prepare instructions for one or more stacks or other components of the AV 402.
Ride-hailing platform 460 can interact with a customer of a ride-hailing service via a ride-hailing application 472 executing on the client computing device 470. The client computing device 470 can be any type of computing system such as, for example and without limitation, a server, desktop computer, laptop computer, tablet computer, smartphone, smart wearable device (e.g., smartwatch, smart eyeglasses or other Head-Mounted Display (HMD), smart ear pods, or other smart in-ear, on-ear, or over-ear device, etc.), gaming system, or any other computing device for accessing the ride-hailing application 472. The client computing device 470 can be a customer's mobile computing device or a computing device integrated with the AV 402 (e.g., the local computing device 410). The ride-hailing platform 460 can receive requests to pick up or drop off from the ride-hailing application 472 and dispatch the AV 402 for the trip.
Map management platform 462 can provide a set of tools for the manipulation and management of geographic and spatial (geospatial) and related attribute data. The data management platform 452 can receive LIDAR point cloud data, image data (e.g., still image, video, etc.), RADAR data, GPS data, and other sensor data (e.g., raw data) from one or more AVs 402, Unmanned Aerial Vehicles (UAVs), satellites, third-party mapping services, and other sources of geospatially referenced data. The raw data can be processed, and map management platform 462 can render base representations (e.g., tiles (2D), bounding volumes (3D), etc.) of the AV geospatial data to enable users to view, query, label, edit, and otherwise interact with the data. Map management platform 462 can manage workflows and tasks for operating on the AV geospatial data. Map management platform 462 can control access to the AV geospatial data, including granting or limiting access to the AV geospatial data based on user-based, role-based, group-based, task-based, and other attribute-based access control mechanisms. Map management platform 462 can provide version control for the AV geospatial data, such as to track specific changes that (human or machine) map editors have made to the data and to revert changes when necessary. Map management platform 462 can administer release management of the AV geospatial data, including distributing suitable iterations of the data to different users, computing devices, AVs, and other consumers of HD maps. Map management platform 462 can provide analytics regarding the AV geospatial data and related data, such as to generate insights relating to the throughput and quality of mapping tasks.
In some embodiments, the map viewing services of map management platform 462 can be modularized and deployed as part of one or more of the platforms and systems of the data center 450. For example, the AI/ML platform 454 may incorporate the map viewing services for visualizing the effectiveness of various object detection or object classification models, the simulation platform 456 may incorporate the map viewing services for recreating and visualizing certain driving scenarios, the remote assistance platform 458 may incorporate the map viewing services for replaying traffic incidents to facilitate and coordinate aid, the ride-hailing platform 460 may incorporate the map viewing services into the client application 472 to enable passengers to view the AV 402 in transit en route to a pick-up or drop-off location, and so on.
While the autonomous vehicle 402, the local computing device 410, and the autonomous vehicle environment 400 are shown to include certain systems and components, one of ordinary skill will appreciate that the autonomous vehicle 402, the local computing device 410, and/or the autonomous vehicle environment 400 can include more or fewer systems and/or components than those shown in
In
Neural network 500 is a multi-layer neural network of interconnected nodes. Each node can represent a piece of information. Information associated with the nodes is shared among the different layers and each layer retains information as information is processed. In some cases, the neural network 500 can include a feed-forward network, in which case there are no feedback connections where outputs of the network are fed back into itself. In some cases, the neural network 500 can include a recurrent neural network, which can have loops that allow information to be carried across nodes while reading in input.
Information can be exchanged between nodes through node-to-node interconnections between the various layers. Nodes of the input layer 520 can activate a set of nodes in the first hidden layer 522a. For example, as shown, each of the input nodes of the input layer 520 is connected to each of the nodes of the first hidden layer 522a. The nodes of the first hidden layer 522a can transform the information of each input node by applying activation functions to the input node information. The information derived from the transformation can then be passed to and can activate the nodes of the next hidden layer 522b, which can perform their own designated functions. Example functions include convolutional, up-sampling, data transformation, and/or any other suitable functions. The output of the hidden layer 522b can then activate nodes of the next hidden layer, and so on. The output of the last hidden layer 522n can activate one or more nodes of the output layer 521, at which an output is provided. In some cases, while nodes in the neural network 500 are shown as having multiple output lines, a node can have a single output and all lines shown as being output from a node represent the same output value.
In some cases, each node or interconnection between nodes can have a weight that is a set of parameters derived from the training of the neural network 500. Once the neural network 500 is trained, it can be referred to as a trained neural network, which can be used to classify one or more activities. For example, an interconnection between nodes can represent a piece of information learned about the interconnected nodes. The interconnection can have a tunable numeric weight that can be tuned (e.g., based on a training dataset), allowing the neural network 500 to be adaptive to inputs and able to learn as more and more data is processed.
The neural network 500 is pre-trained to process the features from the data in the input layer 520 using the different hidden layers 522a, 522b, through 522n in order to provide the output through the output layer 521.
In some cases, the neural network 500 can adjust the weights of the nodes using a training process called backpropagation. A backpropagation process can include a forward pass, a loss function, a backward pass, and a weight update. The forward pass, loss function, backward pass, and parameter/weight update is performed for one training iteration. The process can be repeated for a certain number of iterations for each set of training data until the neural network 500 is trained well enough so that the weights of the layers are accurately tuned.
To perform training, a loss function can be used to analyze error in the output. Any suitable loss function definition can be used, such as a Cross-Entropy loss. Another example of a loss function includes the mean squared error (MSE), defined as E_total=Σ(½ (target−output){circumflex over ( )}2). The loss can be set to be equal to the value of E_total.
The loss (or error) will be high for the initial training data since the actual values will be much different than the predicted output. The goal of training is to minimize the amount of loss so that the predicted output is the same as the training output. The neural network 500 can perform a backward pass by determining which inputs (weights) most contributed to the loss of the network, and can adjust the weights so that the loss decreases and is eventually minimized.
The neural network 500 can include any suitable deep network. One example includes a Convolutional Neural Network (CNN), which includes an input layer and an output layer, with multiple hidden layers between the input and out layers. The hidden layers of a CNN include a series of convolutional, nonlinear, pooling (for downsampling), and fully connected layers. The neural network 500 can include any other deep network other than a CNN, such as an autoencoder, Deep Belief Nets (DBNs), Recurrent Neural Networks (RNNs), among others.
As understood by those of skill in the art, machine-learning based classification techniques can vary depending on the desired implementation. For example, machine-learning classification schemes can utilize one or more of the following, alone or in combination: hidden Markov models; RNNs; CNNs; deep learning; Bayesian symbolic methods; Generative Adversarial Networks (GANs); support vector machines; image registration methods; and applicable rule-based systems. Where regression algorithms are used, they may include but are not limited to: a Stochastic Gradient Descent Regressor, a Passive Aggressive Regressor, etc.
Machine learning classification models can also be based on clustering algorithms (e.g., a Mini-batch K-means clustering algorithm), a recommendation algorithm (e.g., a Minwise Hashing algorithm, or Euclidean Locality-Sensitive Hashing (LSH) algorithm), and/or an anomaly detection algorithm, such as a local outlier factor. Additionally, machine-learning models can employ a dimensionality reduction approach, such as, one or more of: a Mini-batch Dictionary Learning algorithm, an incremental Principal Component Analysis (PCA) algorithm, a Latent Dirichlet Allocation algorithm, and/or a Mini-batch K-means algorithm, etc.
In some embodiments, computing system 600 is a distributed system in which the functions described in this disclosure can be distributed within a datacenter, multiple data centers, a peer network, etc. In some embodiments, one or more of the described system components represents many such components each performing some or all of the function for which the component is described. In some embodiments, the components can be physical or virtual devices.
Example system 600 includes at least one processing unit (Central Processing Unit (CPU) or processor) 610 and connection 605 that couples various system components including system memory 615, such as Read-Only Memory (ROM) 620 and Random-Access Memory (RAM) 625 to processor 610. Computing system 600 can include a cache of high-speed memory 612 connected directly with, in close proximity to, or integrated as part of processor 610.
Processor 610 can include any general-purpose processor and a hardware service or software service, such as services 632, 634, and 636 stored in storage device 630, configured to control processor 610 as well as a special-purpose processor where software instructions are incorporated into the actual processor design. Processor 610 may essentially be a completely self-contained computing system, containing multiple cores or processors, a bus, memory controller, cache, etc. A multi-core processor may be symmetric or asymmetric.
To enable user interaction, computing system 600 includes an input device 645, which can represent any number of input mechanisms, such as a microphone for speech, a touch-sensitive screen for gesture or graphical input, keyboard, mouse, motion input, speech, etc. Computing system 600 can also include output device 635, which can be one or more of a number of output mechanisms known to those of skill in the art. In some instances, multimodal systems can enable a user to provide multiple types of input/output to communicate with computing system 600. Computing system 600 can include communications interface 640, which can generally govern and manage the user input and system output. The communication interface may perform or facilitate receipt and/or transmission wired or wireless communications via wired and/or wireless transceivers, including those making use of an audio jack/plug, a microphone jack/plug, a Universal Serial Bus (USB) port/plug, an Apple® Lightning® port/plug, an Ethernet port/plug, a fiber optic port/plug, a proprietary wired port/plug, a BLUETOOTH® wireless signal transfer, a BLUETOOTH® low energy (BLE) wireless signal transfer, an IBEACON® wireless signal transfer, a Radio-Frequency Identification (RFID) wireless signal transfer, Near-Field Communications (NFC) wireless signal transfer, Dedicated Short Range Communication (DSRC) wireless signal transfer, 802.11 Wi-Fi® wireless signal transfer, Wireless Local Area Network (WLAN) signal transfer, Visible Light Communication (VLC) signal transfer, Worldwide Interoperability for Microwave Access (WiMAX), Infrared (IR) communication wireless signal transfer, Public Switched Telephone Network (PSTN) signal transfer, Integrated Services Digital Network (ISDN) signal transfer, 3G/4G/5G/LTE cellular data network wireless signal transfer, ad-hoc network signal transfer, radio wave signal transfer, microwave signal transfer, infrared signal transfer, visible light signal transfer signal transfer, ultraviolet light signal transfer, wireless signal transfer along the electromagnetic spectrum, or some combination thereof.
Communication interface 640 may also include one or more Global Navigation Satellite System (GNSS) receivers or transceivers that are used to determine a location of the computing system 600 based on receipt of one or more signals from one or more satellites associated with one or more GNSS systems. GNSS systems include, but are not limited to, the US-based Global Positioning System (GPS), the Russia-based Global Navigation Satellite System (GLONASS), the China-based BeiDou Navigation Satellite System (BDS), and the Europe-based Galileo GNSS. There is no restriction on operating on any particular hardware arrangement, and therefore the basic features here may easily be substituted for improved hardware or firmware arrangements as they are developed.
Storage device 630 can be a non-volatile and/or non-transitory and/or computer-readable memory device and can be a hard disk or other types of computer readable media which can store data that are accessible by a computer, such as magnetic cassettes, flash memory cards, solid state memory devices, digital versatile disks, cartridges, a floppy disk, a flexible disk, a hard disk, magnetic tape, a magnetic strip/stripe, any other magnetic storage medium, flash memory, memristor memory, any other solid-state memory, a Compact Disc (CD) Read Only Memory (CD-ROM) optical disc, a rewritable CD optical disc, a Digital Video Disk (DVD) optical disc, a Blu-ray Disc (BD) optical disc, a holographic optical disk, another optical medium, a Secure Digital (SD) card, a micro SD (microSD) card, a Memory Stick® card, a smartcard chip, a EMV chip, a Subscriber Identity Module (SIM) card, a mini/micro/nano/pico SIM card, another Integrated Circuit (IC) chip/card, Random-Access Memory (RAM), Atatic RAM (SRAM), Dynamic RAM (DRAM), Read-Only Memory (ROM), Programmable ROM (PROM), Erasable PROM (EPROM), Electrically Erasable PROM (EEPROM), flash EPROM (FLASHEPROM), cache memory (L1/L2/L3/L4/L5/L #), Resistive RAM (RRAM/ReRAM), Phase Change Memory (PCM), Spin Transfer Torque RAM (STT-RAM), another memory chip or cartridge, and/or a combination thereof.
Storage device 630 can include software services, servers, services, etc., that when the code that defines such software is executed by the processor 610, it causes the system 600 to perform a function. In some embodiments, a hardware service that performs a particular function can include the software component stored in a computer-readable medium in connection with the necessary hardware components, such as processor 610, connection 605, output device 635, etc., to carry out the function.
Embodiments within the scope of the present disclosure may also include tangible and/or non-transitory computer-readable storage media or devices for carrying or having computer-executable instructions or data structures stored thereon. Such tangible computer-readable storage devices can be any available device that can be accessed by a general purpose or special purpose computer, including the functional design of any special purpose processor as described above. By way of example, and not limitation, such tangible computer-readable devices can include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other device which can be used to carry or store desired program code in the form of computer-executable instructions, data structures, or processor chip design. When information or instructions are provided via a network or another communications connection (either hardwired, wireless, or combination thereof) to a computer, the computer properly views the connection as a computer-readable medium. Thus, any such connection is properly termed a computer-readable medium. Combinations of the above should also be included within the scope of the computer-readable storage devices.
Computer-executable instructions include, for example, instructions and data which cause a general-purpose computer, special purpose computer, or special purpose processing device to perform a certain function or group of functions. Computer-executable instructions also include program modules that are executed by computers in stand-alone or network environments. Generally, program modules include routines, programs, components, data structures, objects, and the functions inherent in the design of special-purpose processors, etc. that perform tasks or implement abstract data types. Computer-executable instructions, associated data structures, and program modules represent examples of the program code means for executing steps of the methods disclosed herein. The particular sequence of such executable instructions or associated data structures represents examples of corresponding acts for implementing the functions described in such steps.
Other embodiments of the disclosure may be practiced in network computing environments with many types of computer system configurations, including personal computers, hand-held devices, multi-processor systems, microprocessor-based or programmable consumer electronics, network Personal Computers (PCs), minicomputers, mainframe computers, and the like. Embodiments may also be practiced in distributed computing environments where tasks are performed by local and remote processing devices that are linked (either by hardwired links, wireless links, or by a combination thereof) through a communications network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.
Illustrative examples of the disclosure include:
Aspect 1. A computer-implemented method comprising: accessing data for training a model; separating the data into a first subset of data and a second subset of data; training the model with the first subset of data to generate a trained model; identifying one or more errors associated with running the trained model on the second subset of data; filtering the second subset of data to generate filtered data associated with the one or more errors; and further training the trained model based on the filtered data to generate a refined trained model.
Aspect 2. The computer-implemented method of Aspect 1, further comprising: selecting a specific type of error capable of occurring in applying the model; and mining for the specific type of error using an error miner to identify the one or more errors associated with running the trained model on the second subset of data, the one or more errors being the specific type of error.
Aspect 3. The computer-implemented method of Aspect 2, wherein the error miner is specifically designed to detect the specific type of error.
Aspect 4. The computer-implemented method of any of Aspects 2 and 3, wherein the error miner comprises a filter configured to filter the second subset of data to generate the filtered data associated with the specific type of error.
Aspect 5. The computer-implemented method of Aspect 4, wherein the filtered data associated with the specific type of error that is filtered from the second subset of data comprises data that presents a mode of the specific type of error.
Aspect 6. The computer-implemented method of any of Aspects 4 and 5, wherein the error miner is trained based on replay analysis of execution of the trained model on the data that presents the mode of the specific type of error.
Aspect 7. The computer-implemented method of any of Aspects 1 through 6, further comprising: re-running the trained model on the filtered data associated with the one or more errors as part of performing replay test analysis on the filtered data to identify portions of the second subset of data that actually present the one or errors; and further training the trained model based on the portions of the second subset of data that actually present the one or more errors; and
Aspect 8. The computer-implemented method of any of Aspects 1 through 7, wherein the data for training the model is gathered by an autonomous vehicle (AV) in a real-world environment and the model is part of a software stack for controlling operation of the AV in the real-world environment.
Aspect 9. The computer-implemented method of Aspect 8, wherein the one or more errors include either the model failing to produce a bounding box when an object existed in scene data gathered by the AV or the model producing a bounding when the object did not exist in the scene data.
Aspect 10. A system comprising: one or more processors; and at least one computer-readable storage medium having stored therein instructions which, when executed by the one or more processors, cause the one or more processors to: access data for training a model; separate the data into a first subset of data and a second subset of data; train the model with the first subset of data to generate a trained model; identify one or more errors associated with running the trained model on the second subset of data; filter the second subset of data to generate filtered data associated with the one or more errors; and further train the trained model based on the filtered data to generate a refined trained model.
Aspect 11. The system of Aspect 10, wherein the instructions further cause the one or more processors to: select a specific type of error capable of occurring in applying the model; and mine for the specific type of error using an error miner to identify the one or more errors associated with running the trained model on the second subset of data, the one or more errors being the specific type of error.
Aspect 12. The system of Aspect 11, wherein the error miner is specifically designed to detect the specific type of error.
Aspect 13. The system of any of Aspects 11 and 12, wherein the error miner comprises a filter configured to filter the second subset of data to generate the filtered data associated with the specific type of error.
Aspect 14. The system of Aspect 13, wherein the filtered data associated with the specific type of error that is filtered from the second subset of data comprises data that presents a mode of the specific type of error.
Aspect 15. The system of any of Aspects 13 and 14, wherein the error miner is trained based on replay analysis of execution of the trained model on the data that presents the mode of the specific type of error.
Aspect 16. The system of any of Aspects 10 through 15, wherein the instructions further cause the one or more processors to: re-run the trained model on the filtered data associated with the one or more errors as part of performing replay test analysis on the filtered data to identify portions of the second subset of data that actually present the one or errors; and further train the trained model based on the portions of the second subset of data that actually present the one or more errors; and
Aspect 17. The system of any of Aspects 10 through 16, wherein the data for training the model is gathered by an autonomous vehicle (AV) in a real-world environment and the model is part of a software stack for controlling operation of the AV in the real-world environment.
Aspect 18. The system of Aspect 17, wherein the one or more errors include either the model failing to produce a bounding box when an object existed in scene data gathered by the AV or the model producing a bounding when the object did not exist in the scene data.
Aspect 19. A non-transitory computer-readable storage medium storing instructions for causing one or more processors to: access data for training a model; separate the data into a first subset of data and a second subset of data; train the model with the first subset of data to generate a trained model; identify one or more errors associated with running the trained model on the second subset of data; filter the second subset of data to generate filtered data associated with the one or more errors; and further train the trained model based on the filtered data to generate a refined trained model.
Aspect 20. The non-transitory computer-readable storage medium of Aspect 19, wherein the instructions further cause the one or more processors to: select a specific type of error capable of occurring in applying the model; and mine for the specific type of error using an error miner to identify the one or more errors associated with running the trained model on the second subset of data, the one or more errors being the specific type of error.
Aspect 21. A system comprising means for performing a method according to any of Aspects 1 through 9.
The various embodiments described above are provided by way of illustration only and should not be construed to limit the scope of the disclosure. For example, the principles herein apply equally to optimization as well as general improvements. Various modifications and changes may be made to the principles described herein without following the example embodiments and applications illustrated and described herein, and without departing from the spirit and scope of the disclosure.
Claim language or other language in the disclosure reciting “at least one of” a set and/or “one or more” of a set indicates that one member of the set or multiple members of the set (in any combination) satisfy the claim. For example, claim language reciting “at least one of A and B” or “at least one of A or B” means A, B, or A and B. In another example, claim language reciting “at least one of A, B, and C” or “at least one of A, B, or C” means A, B, C, or A and B, or A and C, or B and C, or A and B and C. The language “at least one of” a set and/or “one or more” of a set does not limit the set to the items listed in the set. For example, claim language reciting “at least one of A and B” or “at least one of A or B” can mean A, B, or A and B, and can additionally include items not listed in the set of A and B.
