Perception simulation for improved autonomous vehicle control

Abstract

A system and method for real world autonomous vehicle perception simulation are disclosed. A particular embodiment includes: configuring a sensor noise modeling module to produce simulated sensor errors or noise data with a configured degree, extent, and timing of simulated sensor errors or noise based on a set of modifiable parameters; using the simulated sensor errors or noise data to generate simulated perception data by simulating errors related to constraints of one or more of a plurality of sensors, and by simulating noise in data provided by a sensor processing module corresponding to one or more of the plurality of sensors; and providing the simulated perception data to a motion planning system for the autonomous vehicle.

Description

COPYRIGHT NOTICE

A portion of the disclosure of this patent document contains material that is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the U.S. Patent and Trademark Office patent files or records, but otherwise reserves all copyright rights whatsoever. The following notice applies to the disclosure herein and to the drawings that form a part of this document: Copyright 2016-2020, TuSimple, Inc., All Rights Reserved.

TECHNICAL FIELD

This patent document pertains generally to tools (systems, apparatuses, methodologies, computer program products, etc.) for autonomous driving simulation systems, trajectory planning, vehicle control systems, and autonomous driving systems, and more particularly, but not by way of limitation, to a system and method for perception simulation for improved autonomous vehicle control.

BACKGROUND

An autonomous vehicle is often configured to follow a trajectory based on a computed driving path. The driving path is based in part on perception data received from sensors on the vehicle. The driving path is also based in part on predicted vehicle and traffic behaviors derived from a vehicle trajectory simulation. When conventional vehicle trajectory simulations are performed, the sensor data received from the vehicle is assumed to be accurate. However in the real world, this vehicle sensor data is often subject to errors, calibration inaccuracies, noise, or other problems. As a result, the vehicle trajectory simulation can produce trajectory data that is not consistent with the actual state of the autonomous vehicle. Therefore, the optimal behavior of the autonomous vehicle cannot be achieved.

SUMMARY

A system and method for perception simulation for improved autonomous vehicle control are disclosed herein. Perception data is needed to make a real world decision in an autonomous driving mode. However, perception data is typically not perfect. In a simulated environment established for the purpose of testing a motion planner's robustness, the simulation is expected to be similar, to the maximum extent, to the behavior of the vehicle in the real world. However, perception noise or other inaccuracies is a common imperfection occurring in the real world. In an example embodiment described herein, the perception noise is modeled in the simulation to improve the quality of the test or the simulated result. The perception noise model of an example embodiment is obtained by setting up calibration on the sensors, and by comparing testing perception result data with ground truths. The perception noise model is then applied to the simulation system for a more realistic perception.

BRIEF DESCRIPTION OF THE DRAWINGS

The various embodiments are illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings in which:

FIG. 1 illustrates a first phase for building an autonomous vehicle perception simulation system in an example embodiment;

FIG. 2 illustrates the components of the autonomous vehicle perception simulation system of an example embodiment in a second phase;

FIG. 3 is a process flow diagram illustrating an example embodiment of a system and method for perception simulation for improved autonomous vehicle control; and

FIG. 4 shows a diagrammatic representation of machine in the example form of a computer system within which a set of instructions when executed may cause the machine to perform any one or more of the methodologies discussed herein.

DETAILED DESCRIPTION

In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the various embodiments. It will be evident, however, to one of ordinary skill in the art that the various embodiments may be practiced without these specific details.

A system and method for perception simulation for improved autonomous vehicle control are disclosed herein. Perception data is needed to make a real world decision in an autonomous driving mode. However, perception data is typically not perfect. In a simulated environment established for the purpose of testing a motion planner's robustness, the simulation is expected to be similar, to the maximum extent, to the behavior of the vehicle in the real world. However, perception noise or other inaccuracies is a common imperfection occurring in the real world. In an example embodiment described herein, the perception noise is modeled in the simulation to improve the quality of the test or the simulated result. The perception noise model of an example embodiment is obtained by setting up calibration on the sensors, and by comparing testing perception result data with ground truths. The perception noise model is then applied to the simulation system for a more realistic perception.

As described in various example embodiments, a system and method for real world autonomous vehicle perception simulation are described herein. Referring to FIG. 1, an example embodiment disclosed herein can be implemented in two phases. In a first phase shown in FIG. 1, an autonomous vehicle perception simulation system 120 under calibration can receive real world perception data 101 from an array of perception sensors and output from perception data processing modules that are typically used in an autonomous vehicle or with an autonomous vehicle simulation system. The intent of the first phase is to analyze the output of calibration results from sensors and perception data processing modules to determine the extent of the errors or inaccuracies experienced with the typical array of perception sensors and perception data processing modules. As shown in the example embodiment of FIG. 1, the autonomous vehicle perception simulation system 120 under calibration can receive real world perception data 101 from a variety of perception sensor devices in the form of image data from image generating devices (e.g., cameras), distance data from light amplification by stimulated emission of radiation (laser) devices or light detection and ranging (LIDAR) devices, location data from global positioning system (GPS) devices, data from sound navigation and ranging (sonar) devices, data from radio detection and ranging (radar) devices, and the like. The perception data collected by the perception simulation system 120 reflects truly realistic, real-world perception data that would be transferred to an autonomous vehicle control system or vehicle simulation system in normal operations. The gathered perception data 101 can be received and processed by the autonomous vehicle perception simulation system 120 in a phase 1 calibration mode. The real world perception data 101 can be used by the autonomous vehicle perception simulation system 120 to build a perception simulation model that can simulate perception data errors as described in more detail below.

Referring still to FIG. 1, components of the autonomous vehicle perception simulation system 120 can perform filtering, object segmentation, object classification, and other image processing or data processing functions to isolate vehicle or object presence and activity in the images, sensor data, and other perception data 101 received by the autonomous vehicle perception simulation system 120. The autonomous vehicle perception simulation system 120 can also receive ground truth data 102. In one example embodiment, the ground truth data 102 can correspond to a data representation of the exact actual location and configuration of every object in a particular environment or every object in proximity to the autonomous vehicle. The ground truth data 102 can be generated by empirical measurement or highly sensitive and accurate sensing devices, such as Lidar data, GPS data, drone data, and the like. In another example embodiment, the ground truth data 102 can correspond to a data representation of a simulation of ground truth data produced by a data simulation system 103. In this embodiment, the exact locations and configurations of every object in a particular environment or every object in proximity to the autonomous vehicle can be precisely defined by the simulation system configured to simulate an environment in which an autonomous vehicle may operate. By use of the actual or simulated ground truth data 102 as a reference, the autonomous vehicle perception simulation system 120 can be executed as described in more detail below.

Real world perception sensor devices and sensor data processing modules can experience errors and other inaccuracies in normal operation. These errors and inaccuracies can be modeled or simulated by the autonomous vehicle perception simulation system 120 based on a comparison of the real world perception data 101 and the ground truth data 102 in the calibration phase. As described above, the ground truth data 102 can represent the exact locations and configurations or simulated locations and configurations of every object in a particular environment or every object in proximity to the autonomous vehicle. The perception simulation system 120 can receive the perception data 101 over a pre-determined time period and compare the received data with the corresponding ground truth data 102. The performance and accuracy of each of the perception sensor devices and sensor data processing modules over a pre-determined time period can be tracked and evaluated. In this manner, the errors and other inaccuracies for each of the perception sensor devices and sensor data processing modules can be detected. Based on this error detection, models can be generated to simulate the typical errors produced by each of the sensor devices and processing modules. These models can be generated by the autonomous vehicle perception simulation system 120. As a result, the autonomous vehicle perception simulation system 120 can produce simulated perception data 125 that corresponds to the modeled data produced for each of the perception sensor devices and sensor data processing modules over the pre-determined time period.

In an alternative embodiment, the autonomous vehicle perception simulation system 120 can be configured to generate models to simulate the typical errors produced by vehicle sensor devices and processing modules based only on the receipt of actual or simulated ground truth data 120 without the real world perception data 101. As described above, the real world perception data 101 can be received and processed by the autonomous vehicle perception simulation system 120 in a phase 1 calibration mode. Once calibrated with the real world perception data 101 in the phase 1 calibration mode, the autonomous vehicle perception simulation system 120 can subsequently generate sensor simulation models without the need to further receive the real world perception data 101. In this embodiment, the real-world perception data 101 is not used to generate the simulated perception data 125 directly. Instead, the real-world perception data 101 is used with actual or simulated ground truth data 102 to initially configure the autonomous vehicle perception simulation system 120. Then, the autonomous vehicle perception simulation system 120 can use actual or simulated ground truth data 102 to generate the simulated perception data 125.

The simulated perception data 125 is important for configuring or training a motion planning system for an autonomous vehicle or a vehicle simulation system. Because the simulated perception data 125 can model the errant or inaccurate data provided by typical sensor devices and processing modules, the simulated perception data 125 more closely corresponds to the errant real world data that would be provided from actual perception sensors or processing modules of an actual vehicle. Because the simulated perception data 125 more closely corresponds to the real world, the motion planning system or vehicle simulation configured or trained with the simulated perception data 125 will also more closely correspond to the real world. Thus, the simulated perception data 125 provided by the autonomous vehicle perception simulation system 120 can be used to produce a more effective control system for an autonomous vehicle. In other words, the simulated perception data 125 generated by the autonomous vehicle perception simulation system 120 is indirectly useful for configuring the control systems for an autonomous vehicle. It will be apparent to those of ordinary skill in the art that the autonomous vehicle perception simulation system 120 and the simulated perception data 125 as described and claimed herein can be implemented, configured, processed, and used in a variety of other applications and systems as well.

Referring now to FIG. 2, a diagram illustrates the components of the autonomous vehicle perception simulation system 120 of an example embodiment. In the example embodiment, the autonomous vehicle perception simulation system 120 can be configured to include an autonomous vehicle perception simulation module 130 configured for execution by a data processor 171 in a computing environment of the autonomous vehicle perception simulation system 120. In the example embodiment, the autonomous vehicle perception simulation module 130 can be configured to include a physical constraint modeling module 173 and a sensor and algorithmic noise modeling module 175. The physical constraint modeling module 173 and the sensor and algorithmic noise modeling module 175 can be configured as software modules for execution by the data processor 171. As described in more detail below, the physical constraint modeling module 173 and the sensor and algorithmic noise modeling module 175 serve to model different types of errors or anomalies occurring in the real world perception data 101.

As shown in FIG. 2, a data storage device or memory 172 can also be provided in the autonomous vehicle perception simulation system 120 of an example embodiment. The memory 172 can be implemented with standard data storage devices (e.g., flash memory, DRAM, SIM cards, or the like) or as cloud storage in a networked server. In an example embodiment, the memory 172 can be used to store a set of sensor configuration data 174. The sensor configuration data 174 corresponds to a data representation of various types of configuration parameters or settings to configure the operation of the various perception sensors of an autonomous vehicle or a vehicle simulation. The memory 172 can also be used to store a set of algorithm performance model data 176. The algorithm performance model data 176 corresponds to a data representation of various types of configuration parameters or settings to configure the operation and performance of the various perception data processing modules used in an autonomous vehicle or a vehicle simulation.

As described above, the autonomous vehicle perception simulation system 120 can produce simulated perception data 125 that corresponds to the modeled data produced for each of the perception sensor devices and sensor data processing modules over the pre-determined time period. Alternatively, the autonomous vehicle perception simulation system 120 can produce simulated perception data 125 that corresponds to actual or simulated ground truth data 102 after calibration with real world perception data 101. In most cases, there are two types of errors or anomalies occurring in the real world perception data 101 that can be modeled by an example embodiment. A first type of error or anomaly corresponds to the physical constraints of the various types of perception sensors used with the autonomous vehicle. For example, some perception sensor devices have a physical range, beyond which the sensor device cannot detect objects or produce useful data. In some cases, the sensor configuration data 174 can be used to quantify these device physical constraints. However, the physical constraints of the sensor devices can vary based on a variety of factors including environmental conditions, vehicle speed, location, or state, the condition or capabilities of the sensor device, and the like. In addition, the context of the usage of the perception sensor is also a factor in the quality of the data produced by the sensor. For example, there may be obstructions or occlusions that may interfere with the ability of a sensor to view a particular location and therefore detect objects or produce useful data related to the particular location. As a result, there are a variety of physical constraint limitations that can occur in the real world perception data 101 that can degrade the quality of the data provided by the perception sensors.

In an example embodiment, the physical constraint modeling module 173 can be configured to model these perception device physical constraint limitations or anomalies. The modeling can be based in part on the initial configuration and calibration of each perception device performed during the phase 1 calibration described above and retained as sensor configuration data 174. The modeling of each perception device can also be based in part on the context in which the vehicle or vehicle simulation is operating. For example, the perception device modeling can be varied based on day operation versus night operation or heavy traffic conditions versus light traffic conditions. Heavy traffic conditions can produce a greater number of occlusion anomalies in comparison to light traffic conditions. Similarly, other contexts and conditions related to the perception device physical constraint limitations can be modeled by the physical constraint modeling module 173. The simulated physical constraint modeling for each perception device can be included in the simulated perception data 125 provided to a vehicle motion planning system 220, an autonomous vehicle control module configuration system, or vehicle simulation system.

In an example embodiment, a second type of error or anomaly occurring in the real world perception data 101 can be modeled by the sensor and algorithmic noise modeling module 175. In most vehicles that include a variety of perception sensor devices, various types of processing modules, algorithms, filters, detectors, and the like are used to pre-process the raw sensor data provided by the sensor devices. For example, image processing modules can be used to process the images received from a camera to detect objects, other vehicles, roadway lane stripes, etc. in the images. Other perception processing modules can include localization or navigation modules that use GPS or inertial data to determine a vehicle location and trajectory or path. Each of these perception data processing modules can be subject to noise, errors, inaccuracies, or other anomalies. The sensor and algorithmic noise modeling module 175 of an example embodiment can be configured to model these sensor data processing errors, noise, and anomalies. The modeling can be based in part on the initial configuration and model data provided for each perception data processing module as performed during the phase 1 calibration described above and retained as algorithm performance model data 176. The modeling of each perception data processing module can also be based in part on the context in which the vehicle or vehicle simulation is operating. The modeling of each perception data processing module can be configured to mimic the inaccuracy of actual sensors and perception devices and their corresponding perception data processing modules. For example, the sensor and algorithmic noise modeling module 175 of an example embodiment can be configured produce false positive or false negative object detections, which could be produced by a perception processing module performing image processing on input image data. Similarly, the sensor and algorithmic noise modeling module 175 can be configured produce noise or inaccuracies in the localization of the vehicle (i.e., the determination of the current location of the vehicle). The sensor and algorithmic noise modeling module 175 can be further configured produce noise or inaccuracies in the lane detection, which can be processed by an image processing module. The sensor and algorithmic noise modeling module 175 can be further configured produce noise or inaccuracies in the classification of detected objects or the estimated locations of detected objects, which can be processed by an image processing module. In a variety of other examples, the sensor and algorithmic noise modeling module 175 can be further configured produce noise or inaccuracies in any of the perception sensor processing modules used in an autonomous vehicle or vehicle simulation. Similarly, the physical constraint modeling module 173, as described above, can be configured produce errors, noise, or inaccuracies in any of the perception sensor devices used in an autonomous vehicle or vehicle simulation. The degree, extent, and timing of the errors or noise introduced into the modeling of the perception sensor devices or the perception sensor processing modules can be configurable based on a set of user-modifiable parameters. The simulated physical constraint modeling for each perception device and the simulated noise or inaccuracies in any of the perception sensor processing modules can be included in the simulated perception data 125 provided to a vehicle motion planning system 220, an autonomous vehicle control module configuration system, or vehicle simulation system.

As a result, the autonomous vehicle perception simulation system 120 can produce simulated perception data 125 that corresponds to the modeled data produced for each of the perception sensor devices and sensor data processing modules over the pre-determined time period. Alternatively, the autonomous vehicle perception simulation system 120 can produce simulated perception data 125 that corresponds to actual or simulated ground truth data 102 after calibration with real world perception data 101. The simulated perception data 125 can be used for configuring or training a motion planning system for an autonomous vehicle or a vehicle simulation system.

Referring now to FIG. 3, a flow diagram illustrates an example embodiment of a system and method 1000 for autonomous vehicle perception simulation. The example embodiment can be configured for: receiving perception data from a plurality of sensors of an autonomous vehicle (processing block 1010); configuring the perception simulation operation based on a comparison of the perception data against ground truth data (processing block 1020); generating simulated perception data by simulating errors related to the physical constraints of one or more of the plurality of sensors, and by simulating noise in data provided by a sensor processing module corresponding to one or more of the plurality of sensors (processing block 1030); and providing the simulated perception data to a motion planning system for the autonomous vehicle (processing block 1040).

FIG. 4 shows a diagrammatic representation of a machine in the example form of a computing system 700 within which a set of instructions when executed and/or processing logic when activated may cause the machine to perform any one or more of the methodologies described and/or claimed herein. In alternative embodiments, the machine operates as a standalone device or may be connected (e.g., networked) to other machines. In a networked deployment, the machine may operate in the capacity of a server or a client machine in server-client network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The machine may be a personal computer (PC), a laptop computer, a tablet computing system, a Personal Digital Assistant (PDA), a cellular telephone, a smartphone, a web appliance, a set-top box (STB), a network router, switch or bridge, or any machine capable of executing a set of instructions (sequential or otherwise) or activating processing logic that specify actions to be taken by that machine. Further, while only a single machine is illustrated, the term “machine” can also be taken to include any collection of machines that individually or jointly execute a set (or multiple sets) of instructions or processing logic to perform any one or more of the methodologies described and/or claimed herein.

The example computing system 700 can include a data processor 702 (e.g., a System-on-a-Chip (SoC), general processing core, graphics core, and optionally other processing logic) and a memory 704, which can communicate with each other via a bus or other data transfer system 706. The mobile computing and/or communication system 700 may further include various input/output (I/O) devices and/or interfaces 710, such as a touchscreen display, an audio jack, a voice interface, and optionally a network interface 712. In an example embodiment, the network interface 712 can include one or more radio transceivers configured for compatibility with any one or more standard wireless and/or cellular protocols or access technologies (e.g., 2nd (2G), 2.5, 3rd (3G), 4th (4G) generation, and future generation radio access for cellular systems, Global System for Mobile communication (GSM), General Packet Radio Services (GPRS), Enhanced Data GSM Environment (EDGE), Wideband Code Division Multiple Access (WCDMA), LTE, CDMA2000, WLAN, Wireless Router (WR) mesh, and the like). Network interface 712 may also be configured for use with various other wired and/or wireless communication protocols, including TCP/IP, UDP, SIP, SMS, RTP, WAP, CDMA, TDMA, UMTS, UWB, WiFi, WiMax, Bluetooth™, IEEE 802.11x, and the like. In essence, network interface 712 may include or support virtually any wired and/or wireless communication and data processing mechanisms by which information/data may travel between a computing system 700 and another computing or communication system via network 714.

The memory 704 can represent a machine-readable medium on which is stored one or more sets of instructions, software, firmware, or other processing logic (e.g., logic 708) embodying any one or more of the methodologies or functions described and/or claimed herein. The logic 708, or a portion thereof, may also reside, completely or at least partially within the processor 702 during execution thereof by the mobile computing and/or communication system 700. As such, the memory 704 and the processor 702 may also constitute machine-readable media. The logic 708, or a portion thereof, may also be configured as processing logic or logic, at least a portion of which is partially implemented in hardware. The logic 708, or a portion thereof, may further be transmitted or received over a network 714 via the network interface 712. While the machine-readable medium of an example embodiment can be a single medium, the term “machine-readable medium” should be taken to include a single non-transitory medium or multiple non-transitory media (e.g., a centralized or distributed database, and/or associated caches and computing systems) that store the one or more sets of instructions. The term “machine-readable medium” can also be taken to include any non-transitory medium that is capable of storing, encoding or carrying a set of instructions for execution by the machine and that cause the machine to perform any one or more of the methodologies of the various embodiments, or that is capable of storing, encoding or carrying data structures utilized by or associated with such a set of instructions. The term “machine-readable medium” can accordingly be taken to include, but not be limited to, solid-state memories, optical media, and magnetic media.

The Abstract of the Disclosure is provided to allow the reader to quickly ascertain the nature of the technical disclosure. It is submitted with the understanding that it will not be used to interpret or limit the scope or meaning of the claims. In addition, in the foregoing Detailed Description, it can be seen that various features are grouped together in a single embodiment for the purpose of streamlining the disclosure. This method of disclosure is not to be interpreted as reflecting an intention that the claimed embodiments require more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive subject matter lies in less than all features of a single disclosed embodiment. Thus, the following claims are hereby incorporated into the Detailed Description, with each claim standing on its own as a separate embodiment.

Claims

1. A system comprising: a data processor; anda perception simulation module, executable by the data processor, the perception simulation module being configured to: configure a sensor noise modeling module to produce simulated sensor errors or noise data with a configured degree, extent, and timing of simulated sensor errors or noise based on a set of modifiable parameters;use the simulated sensor errors or noise data to generate simulated perception data by simulating errors related to constraints of one or more of a plurality of sensors, and by simulating noise in data provided by a sensor processing module corresponding to one or more of the plurality of sensors; andprovide the simulated perception data to a motion planning system for an autonomous vehicle.

2. The system of claim 1 wherein the sensor noise modeling module being configured to produce simulated sensor errors or noise data includes simulating a false positive or false negative object detection.

3. The system of claim 1 wherein the sensor noise modeling module being configured to produce simulated sensor errors or noise data includes simulating noise in vehicle localization.

4. The system of claim 1 wherein the sensor noise modeling module being configured to produce simulated sensor errors or noise data includes simulating noise in lane detection.

5. The system of claim 1 wherein the sensor noise modeling module being configured to produce simulated sensor errors or noise data includes simulating errors in object type classification.

6. The system of claim 1 wherein the sensor noise modeling module being configured to produce simulated sensor errors or noise data includes simulating errors in object location.

7. The system of claim 1 wherein the sensor noise modeling module being configured to produce simulated sensor errors or noise data includes simulating errors in data produced by an image processing module.

8. A method comprising: configuring a sensor noise modeling module to produce simulated sensor errors or noise data with a configured degree, extent, and timing of simulated sensor errors or noise based on a set of modifiable parameters;using the simulated sensor errors or noise data to generate simulated perception data by simulating errors related to constraints of one or more of a plurality of sensors, and by simulating noise in data provided by a sensor processing module corresponding to one or more of the plurality of sensors; andproviding the simulated perception data to a motion planning system for an autonomous vehicle.

9. The method of claim 8 wherein configuring the sensor noise modeling module to produce simulated sensor errors or noise data includes simulating a false positive or false negative result produced by a perception processing module.

10. The method of claim 8 wherein configuring the sensor noise modeling module to produce simulated sensor errors or noise data includes simulating noise produced by a perception sensor device.

11. The method of claim 8 wherein configuring the sensor noise modeling module to produce simulated sensor errors or noise data includes simulating noise in perception sensor devices and sensor data processing modules over a pre-determined time period.

12. The method of claim 8 including producing simulated perception data that corresponds to actual or simulated ground truth data.

13. The method of claim 8 including configuring or training the motion planning system for the autonomous vehicle or a vehicle simulation system.

14. The method of claim 8 wherein configuring the sensor noise modeling module to produce simulated sensor errors or noise data includes simulating errors corresponding to physical constraints of a perception sensor used with the autonomous vehicle.

15. A non-transitory machine-useable storage medium embodying instructions which, when executed by a machine, cause the machine to: configure a sensor noise modeling module to produce simulated sensor errors or noise data with a configured degree, extent, and timing of simulated sensor errors or noise based on a set of modifiable parameters;use the simulated sensor errors or noise data to generate simulated perception data by simulating errors related to constraints of one or more of a plurality of sensors, and by simulating noise in data provided by a sensor processing module corresponding to one or more of the plurality of sensors; andprovide the simulated perception data to a motion planning system for an autonomous vehicle.

16. The non-transitory machine-useable storage medium of claim 15 wherein the plurality of sensors includes: image generating devices, cameras, distance measuring devices, light amplification by stimulated emission of radiation (laser) devices or light detection and ranging (LIDAR) devices, global positioning systems (GPS) devices, sound navigation and ranging (sonar) devices, or radio detection and ranging (radar) devices.

17. The non-transitory machine-useable storage medium of claim 15 wherein the sensor noise modeling module being configured to produce simulated sensor errors or noise data includes simulating a false positive or false negative object classification.

18. The non-transitory machine-useable storage medium of claim 15 wherein the sensor noise modeling module being configured to produce simulated sensor errors or noise data includes simulating noise in vehicle trajectory.

19. The non-transitory machine-useable storage medium of claim 15 wherein the sensor noise modeling module being configured to produce simulated sensor errors or noise data includes simulating noise in roadway detection.

20. The non-transitory machine-useable storage medium of claim 15 being configured to perform a calibration phase based on a comparison of real world perception data and ground truth data.