System and method for correcting orientation errors

Abstract

An orientation system includes an orientation sensor, a distance sensor, and a vehicle processing unit. The orientation sensor is configured to generate orientation data. The distance sensor is configured to generate relative distance data that measures a relative distance to an object external to the vehicle. The vehicle processing unit is configured to receive the orientation data from the orientation sensor and to receive the relative distance data from the distance sensor, wherein the vehicle processing unit detects an orientation error based on the relative distance data.

Description

TECHNICAL FIELD

[0001] The present disclosure relates generally to correcting orientation errors accumulated in orientation instruments, and more particularly to correcting orientation errors with object detection feedback. BACKGROUND

[0002] Vehicles, particularly autonomous vehicles or self-driving vehicles, utilize orientation sensors (sometimes referred to as inertial measurement units or IMUs) to determine the orientation of the vehicle (i.e., roll, pitch, yaw). For example, the orientation sensors can include one or more of an accelerometer, a gyroscope, and a magnetometer to determine and maintain the orientation of the vehicle. The orientation of the vehicle can be utilized by many vehicle systems, including vehicle positioning systems and object detection systems. For example, in some embodiments, orientation information is used to provide dead-reckoning estimates of the vehicle’s position in situations where satellite-based positioning systems are not available. Additionally, the orientation of the vehicle can be used in conjunction with one or more object sensors (e.g., radar-based sensors, LiDAR-based sensors, laser sensors, etc.) to determine the position of objects relative to the vehicle.

[0003] Over time, the orientation sensors can accumulate errors. As the magnitude of these errors increases, it can adversely affect systems that utilize the orientation information. In some embodiments, the orientation errors (at least the orientation errors relative to the vertical axis) can be corrected based on input provided by an auxiliary sensor such as a tiltmeter. However, adding another sensor increases cost and decreases reliability associated with the vehicle. It would be beneficial to develop systems and methods that correct for orientation errors without the need for additional, dedicated sensors. SUMMARY

[0004] According to one aspect, an orientation system is described that includes an orientation sensor, a distance sensor, and a vehicle processing unit. The distance sensor is configured to measure a relative distance to an object outside of the vehicle. The vehicle processing unit is configured to receive orientation data from the orientation sensor and receive relative distance data from the distance sensor, wherein the vehicle processing unit detects an orientation error based on the relative distance data.

[0005] According to another aspect, a method of correcting orientation errors accumulated in an orientation sensor includes receiving orientation data defining an orientation of a vehicle along one or more orientation axes. The method further includes receiving relative distance data from a distance sensor, where the relative distance data includes a measured relative distance to an external object. A planar surface is detected in the received relative distance data, and one or more surface axes associated with the detected planar surface are detected, and a variance between the orientation axes and the planar axes is calculated. One or more of the orientation axes are corrected in response to the calculated variance being less than a threshold value. BRIEF DESCRIPTION OF DRAWINGS

[0006] Figure 1 is a block diagram of components for correcting orientation errors based on relative distances measured by a distance sensor according to some embodiments.

[0007] Figure 2a and Figure 2b is a top view and side view of a vehicle including multiple position sensors for correcting orientation errors according to some embodiments.

[0008] Figure 3 is a block diagram of multiple LiDAR sensors for correcting orientation errors accumulated in an IMU according to some embodiments.

[0009] Figure 4 is a flowchart showing steps performed to correct orientation errors based on relative distances measured by a distance sensor according to some embodiments.

[0010] Figure 5 is a diagram visually showing correction of orientation axes based on a detected planar surface according to some embodiments. DETAILED DESCRIPTION

[0011] Orientation sensors are used on vehicles to provide on-board systems with information related to the orientation of the vehicle. Over time, these systems can accumulate errors that degrade the overall performance of the vehicle. According to some aspects of the present disclosure, relative distance data received from one or more distance sensors is used to correct errors accumulated in the orientation sensor. More specifically, the relative distance data is analyzed to detect planar surfaces (e.g., sides of buildings, road signs, etc.). For those planar surfaces that are approximately aligned with an orientation axis of the vehicle, it is assumed that these surfaces are aligned with the vehicle axis. For example, a building located near the vehicle likely has a planar surface that is approximately perpendicular to the ground. The perpendicular axis defined by the side of the building is likely aligned with the perpendicular axis of the vehicle near the building. The difference between the perpendicular axis (or horizontal axis) defined by the planar surface and the orientation axis provided by the orientation sensor is determined to represent an error in the orientation data. Responsive to the detected change or difference between the respective axes, an adjustment or correction is provided to modify the orientation of the orientation sensor to correct for the error. Although the assumption that the axis defined by the planar surface is correctly aligned with the vehicle axis is clearly incorrect on a case-by-case basis (e.g., the side of a building, a road sign, etc. is slightly off perpendicular), the error or change in the axis defined by the planar surface relative to the vehicle is Gaussian. That is, the likelihood of a building or road sign being off perpendicular in one direction is as likely as being off perpendicular in the opposite direction. As a result, the cumulative corrections over time and multiple planar surfaces will result in the orientation being corrected toward approximately correct values. In this way, distance sensors that are typically employed on vehicles can be employed to prevent orientation errors from accumulating in the corresponding orientation sensor.

[0012] Figure 1 is a block diagram of various sensors (including distance sensors 102 and orientation sensors 104) configured to communicate with a vehicle processing unit 106. The orientation sensors 104 include one or more of an accelerometer, a gyroscope, a magnetometer, and / or other sensors used to measure and determine the orientation of the vehicle. The vehicle processing unit 106 utilizes the orientation data in many applications, such as determining the heading of the vehicle, interpreting objects detected by one or more distance sensors 102, and / or determining the location of the vehicle. As discussed above, the orientation sensors 104 can accumulate errors over time that, if unchecked, can result in false orientation estimates being provided to the vehicle processing unit 106.

[0013] The one or more distance sensors 102 are used to measure the relative distance to objects located outside the vehicle. The one or more distance sensors 102 can include one or more of laser distance and ranging (LiDAR) sensors, camera-based or vision-based sensors, proximity sensors, laser sensors, radar sensors, acoustic sensors, and other sensors. In some embodiments, each of the distance sensors 102 collects a plurality of points (sometimes referred to as a point cloud), where each point in the point cloud describes the distance of that point from the sensor. Subsequent analysis of the point cloud by the vehicle processing unit 106 (or the sensor itself) allows the points to be clustered into surfaces representing objects located near the vehicle 100. According to some embodiments, errors in the orientation sensor 104 are corrected with assumptions about the surfaces detected by the distance sensors 102. In Figure 1 the illustrated embodiment, the vehicle processing unit 106 detects the errors and provides orientation feedback to the orientation sensor 104 to correct for the detected errors. In other embodiments, no feedback is required, but the vehicle processing unit 106 modifies the orientation data provided by the orientation sensor 104 to correct for the detected errors.

[0014] With respect to Figure 2a and Figure 2b , top and side views of a vehicle 200 are shown that includes a plurality of distance sensors 102a-102d. Figure 2a Assumptions relied upon to correct the vertical axis of the orientation sensor are shown, and Figure 2b Assumptions relied upon to correct the horizontal axis of the orientation sensor are shown. The orientation of the vehicle is represented by rotations about three principal axes labeled here as x, y, and z (referred to as vehicle orientation). For purposes of discussion, the z-axis is oriented vertically, the x-axis is oriented longitudinally along the length of the vehicle 200, and the y-axis is oriented laterally along the width of the vehicle 200. The orientation sensor 104 (shown in Figure 1 ) is configured to measure the orientation of the vehicle 200 and provide orientation data to the vehicle processing unit 106 (also shown in Figure 1 ). The one or more distance sensors 102a-102d are located at various positions around the vehicle and are configured to detect objects located near or outside the vehicle 200.

[0015] Figure 2ais a side view of the vehicle 200 showing an object detected by the distance sensor 102b that is vertically aligned. Specifically, the sensor 102b measures the relative distance between the sensor 102b and an object located within the field of view of the sensor 102b. The set of relative distances is analyzed to detect surfaces, such as those associated with the stop sign 204 and the road sign 206. In some embodiments, planar surfaces detected by the distance sensor 102 are used to detect orientation errors. For example, in Figure 2a In the illustrated embodiment, the road sign 206 provides a planar surface that can be selected as useful for detecting orientation errors. For a vertically oriented planar surface, the assumption is made that the vertical axis is approximately oriented upright upward and downward (i.e., perpendicular to the horizontal) and thus aligned with the vertical axis of the vehicle. Changes or errors detected by comparing the vertical axis z' associated with the planar surface to the vertical axis defined by the orientation sensor 106 are attributed to errors in the orientation sensor 106. In response to the detected changes, an adjustment or correction value can be generated to correct the orientation indicated by the orientation sensor 106. Although this assumption can not be valid in certain situations, it is further assumed that errors in the axis defined by the planar surface (such as the z' axis) relative to the vehicle axis are Gaussian in nature and cancel out as the number of planar surfaces analyzed increases. That is, a first road sign 106 that is slightly misaligned in a first direction from the vertical direction (and thus from the vertical axis z of the vehicle 200) will be canceled out by a second or subsequent road sign that is misaligned in the opposite direction. In addition to Figure 2a In addition to the road signs used as examples in the description, most structures having planar surfaces that are elevated in the vertical direction are approximately perpendicular to the horizon - and thus represent a vertical axis z' that should be aligned with the vertical axis z of the vehicle. For example, the planar surfaces associated with most buildings are approximately vertical in nature. Although this can not be true for all buildings, in most cases buildings are oriented approximately vertically with errors relative to true vertical orientation that are Gaussian in nature. In addition, as discussed in more detail below, if the axis defined by a planar surface is significantly different from the orientation axis of the vehicle (e.g., a road sign is partially knocked over, extends at a 45 degree angle relative to the horizontal), steps can be taken to discard that planar surface from the analysis.

[0016] Similar analysis can be provided with respect to horizontal axes (e.g., x, y) in addition to the vertical axis, as illustrated in the example Figure 2a Figure 2b ​In the illustrated embodiment, distance sensor 102d detects the side of building 210. Building 210 is relatively flat and defines an axis x'. It is assumed that the horizontal axis x' is approximately aligned with the longitudinal axis x of vehicle 200. The difference or change between the horizontal axis x' defined by planar surface 210 and the longitudinal orientation provided by the orientation sensor can be attributed to an error in the orientation sensor 104. The assumption provided with respect to the horizontal axis is that most roads are aligned (parallel or perpendicular) with adjacent buildings. As a result, the horizontal axis x', y' defined by planar surfaces in the vicinity of vehicle 200 is likely to be aligned with the orientation axis x, y of vehicle 200. In some embodiments, an additional assumption can be relied upon, namely that if a building is not aligned with the road, but is aligned with a cardinal direction (i.e., north, south, east, west), the building can be used to detect an error in the orientation sensor.

[0017] Reference is now made to Figure 3 , a block diagram illustrating components included in orientation system 300, in accordance with some embodiments. In Figure 3 In the illustrated embodiment, orientation system 300 includes a plurality of LiDAR sensors 302, 302b,..., 302N, a plurality of LiDAR controllers 304a, 304b,..., 304N, an IMU 306, an IMU correction unit 308, and a vehicle processing unit 310. The plurality of LiDAR sensors 302, 302b,..., 302N are used to collect relative distance data, which is provided to the plurality of LiDAR controllers 304a, 304b,..., 304N, respectively. In this embodiment, the LiDAR controllers (generally referred to as LiDAR controllers 304) process the point clouds collected by the plurality of LiDAR sensors 302. In some embodiments, the LiDAR controllers 304 cluster points and detect surfaces. Vehicle processing unit 310 receives the detected surfaces from the LiDAR controllers 304 and utilizes these surfaces to detect and identify objects (provided as output labeled "LiDAR detected objects"). In addition, vehicle processing unit 310 can use the surfaces detected by the LiDAR controllers 304 to detect errors in the orientation data provided by the IMU 306.

[0018] In some embodiments, vehicle processing unit 310 looks at the surfaces provided by the plurality of LiDAR controllers 304 and detects planar surfaces to compare with the orientation data received from the IMU. In some embodiments, a surface is identified as a planar surface if a plurality of points located between a first point and a second point lie on the same plane. For example, Figure 5A surface 500 detected by the LiDAR controller 304 is shown. A number of points located on the surface 500 are selected and used to determine whether the points extend along a single axis or a single plane. In some embodiments, the determination of whether a surface is planar is based on a smaller number of points than are included in the surface. In some embodiments, points are selected along the entire surface 500, while in other embodiments, points are selected along an axis defined by the surface 500. For example, as shown in Figure 5 axis 504 is defined by a first point 508a selected at or near the bottom of the surface 500 and a second point 508b selected at or near the top of the surface 500. If a number of points (labeled as "510") located between the first point 508a and the second point 508b are located on approximately the same plane, then the surface 500 is determined to be planar. In some embodiments, a confidence level is assigned to the detected planar surface 500 based on how closely the selected points are aligned along the same plane. In some embodiments, the confidence level must be above a threshold value, otherwise the surface will be considered non-planar and discarded. In some embodiments, the confidence level is retained and used in subsequent steps to weight corrections or adjustments made to the orientation data.

[0019] Assuming the surface is determined to be planar, one or more axes that define the planar surface are compared to one or more axes of the orientation data provided by the IMU 306. In some embodiments, if the difference or change between the compared axes is greater than a threshold value, then this indicates that the planar surface 500 is not closely enough aligned with the orientation axes of the IMU 306 to be used for correction. The assumption relied on here is that the orientation data provided by the IMU 306 can have some error, but that the error will not be too large. For example, if a stop sign is tilted at a 45 degree angle, then the planar surface of the sign will be sufficiently flat, but the difference between the vertical orientation axis provided by the IMU 306 and the axis defined by the sign will be too large for the sign to be aligned with the vertical axis of the vehicle.

[0020] In some embodiments, if the change between the compared axes is greater than a threshold value, then a second check can be performed to determine whether the axis defined by the planar surface is aligned with a cardinal direction. In some embodiments, this check is performed only with respect to the horizontal axes. If the change or difference between the axis associated with the planar surface and a cardinal axis is greater than a threshold value, then it is determined that the planar surface cannot be used to correct the orientation data and the surface is discarded.

[0021] Assuming a comparison between one or more axes associated with a planar surface and a cardinal axis or an orientation axis defined by the IMU 306, one or more axes defined by the planar surface are assumed to be aligned with one or more axes of the vehicle. Based on this assumption, a change or difference between the one or more axes defined by the planar surface and the one or more axes provided by the IMU 306 is determined to be a result of an error of the IMU 306. For example, in the embodiment shown in FIG. 5, the planar surface 500 is defined by a vertical axis 504 that is compared to a vertical axis 502 defined by the IMU 306. A change between the two axes is defined by an angle 506, which represents an error attributable to the IMU 306. Figure 5

[0022] Based on the detected change between the respective axes, the vehicle processing unit 310 generates orientation feedback provided to the IMU correction unit 308. In some embodiments, a magnitude of the orientation feedback is based on a magnitude of the change detected between the respective axes. In some embodiments, if the change is greater than a threshold value (referred to herein as a“nudge threshold”), a predetermined value is assigned to the magnitude of the orientation feedback. In some embodiments, if the change is greater than the nudge threshold, the magnitude of the orientation feedback is equal to the nudge threshold. In other embodiments, the magnitude of the orientation feedback can be greater or less than the nudge threshold. In some embodiments, if the change is less than the nudge threshold, the magnitude of the orientation feedback is equal to the magnitude of the change. In some embodiments, the magnitude of the orientation feedback provided to the IMU correction unit 308 is further based on a confidence level associated with the determination. For example, this can include a confidence level associated with the planar surface detected by the LiDAR sensor 302, where a higher confidence level results in the orientation feedback being assigned a greater value than a lower confidence level of the detected planar surface. In some embodiments, the orientation feedback provided to the IMU correction unit 308 is added to previously calculated orientation feedback such that the IMU correction unit 308 accumulates the orientation feedback values calculated by the vehicle processing unit 310. In some embodiments, the orientation feedback provided to the IMU correction unit 308 is latched to ensure that the correction values are retained by the IMU correction unit 308.

[0023] In this way, the orientation data provided by the IMU 306 is augmented by the correction values stored by the IMU correction unit 308. Each time orientation data is calculated by the IMU 306, the orientation data is adjusted by the values stored in the IMU correction unit 308, which are continually updated based on a comparison between the (corrected) IMU orientation data and the planar surfaces detected by the one or more LiDAR sensors 302 and the LiDAR controller 304.

[0024] ​Figure 4 is a flowchart showing steps performed to correct orientation data based on relative distances collected by one or more distance sensors. It should be understood that the order of the steps can be modified, and in some embodiments, one or more of the steps can be omitted.

[0025] At step 401, orientation data is received from an orientation sensor (e.g., an IMU sensor). At step 402, relative distance data is received from one or more distance sensors. As described above, many different types of distance sensors can be utilized, including one or more of LiDAR-based sensors, radar-based sensors, vision-based sensors, etc. In some embodiments, the relative distance data collected by the distance sensors is provided in the form of a point cloud, where each point is defined by a distance from the sensor to an object.

[0026] At step 404, surfaces are detected within the received relative distance data. The processing of the relative distance data can be performed locally by the distance sensors, or separately from the distance sensors, such as by a vehicle processing unit. The step of detecting surfaces within the relative distance data can be utilized in other applications, such as as part of an object detection process. In some applications, the step of detecting surfaces is referred to as point clustering, where multiple points included in a point cloud are clustered together based on their distance from one another to form and detect surfaces. In some embodiments, additional analysis is performed on the detected surfaces to classify the surfaces as objects (e.g., cars, buildings, pedestrians, etc.).

[0027] At step 406, the surfaces detected at step 404 are analyzed to determine whether they are planar. In some embodiments, the determination of whether an object is planar involves selecting multiple points along the detected surface and checking to determine whether the multiple points are positioned along a particular plane or axis. In Figure 5 In the example shown, a surface 500 has been detected, multiple points are selected relative to the surface 500, and are used to determine whether the surface is planar. In some embodiments, as shown, the selected points are positioned along an axis, but in other embodiments, the points can be selected randomly along the entire surface. Additionally, the number of points tested to determine whether a surface is planar can be less than the total number of points identified relative to the surface. If the surface is identified as non-planar at step 406, the surface is discarded at step 408. If the surface is identified as planar at step 406, the process continues at step 410. Figure 5 In the example shown, a surface 500 has been detected, multiple points are selected relative to the surface 500, and are used to determine whether the surface is planar. In some embodiments, as shown, the selected points are positioned along an axis, but in other embodiments, the points can be selected randomly along the entire surface. Additionally, the number of points tested to determine whether a surface is planar can be less than the total number of points identified relative to the surface. If the surface is identified as non-planar at step 406, the surface is discarded at step 408. If the surface is identified as planar at step 406, the process continues at step 410.

[0028] In some embodiments, at step 410, a confidence level is assigned to the surface based on the analysis of whether the surface is planar. In some embodiments, a higher confidence level is assigned to surfaces that are determined to be more planar, and a lower confidence level is assigned to surfaces that are determined to be less planar (assuming the surface is sufficiently planar so as not to be discarded entirely). Additionally, the confidence level can also be based on the number of points relative to the surface that were analyzed, where the confidence associated with a surface being planar increases as the number of points analyzed increases (assuming the points are determined to be positioned along an axis or plane). As described in more detail below, in some embodiments, the confidence level assigned to a surface is used to determine an adjustment or correction applied to the orientation sensor, where a high confidence that a surface is planar results in a larger adjustment / correction.

[0029] In some embodiments, at step 412, the orientation of the detected planar surface is translated to the orientation of the vehicle. For example, if the distance sensor used to capture the planar surface is oriented in a direction that is known to be offset from the orientation of the vehicle or a particular axis (i.e., downward), then the surface is translated to account for the offset. Typically, the orientation of the distance sensor relative to the orientation of the vehicle is known, such that the translation relative to each surface detected by a particular sensor remains the same.

[0030] At step 413, a change or difference between the orientation axis of the vehicle and the orientation of the detected planar surface is calculated. The change represents the error between the orientation axis provided by the orientation sensor and the orientation of the planar surface. However, this does not necessarily mean that the change represents an error of the orientation sensor, only that there is a difference between the axis of the orientation axis of the vehicle (provided by the orientation sensor) and the orientation of the planar surface. In some embodiments, at step 414, the change is used to determine whether the planar surface is well suited to estimate the error of the orientation sensor, and subsequently, at steps 416-422, determinations are made regarding how to correct the orientation sensor based on the orientation of the detected planar surface.

[0031] At step 414, the axis difference calculated at step 413 is compared to a threshold. This comparison is used to determine whether the axes are closely enough aligned such that the axis associated with the detected surface is determined to be a good candidate for use in correcting the orientation sensor. Planar surfaces that are not aligned with the axis of the vehicle (e.g., a building aligned at an angle with respect to the street on which the vehicle is located) are not useful for correcting the orientation sensor and should be discarded. In some embodiments, if the change or difference between the axes is less than the threshold, this indicates that the axis associated with the planar surface is likely aligned with the orientation axis of the vehicle and can be used to correct the orientation sensor. In some embodiments, the threshold is a fixed value, while in other embodiments, the threshold can vary. For example, as the duration between corrections increases, the threshold can be increased to accept more planar surfaces that can be useful. That is, in environments where multiple planar surfaces aligned with the orientation of the vehicle are detected, it can be beneficial to discard even a small number of those planar surfaces that are determined to be misaligned with the orientation of the vehicle. In environments where very few planar surfaces are identified as being useful for correcting orientation errors (e.g., driving in the countryside), it can be beneficial to increase the threshold to utilize planar surfaces that would otherwise be discarded.

[0032] In some embodiments, if the change between the axes is greater than the threshold at step 414, the planar surface is discarded at step 408. In other embodiments, if the change between the axes is greater than the threshold, at step 418, the axis or axes associated with the planar surface are compared to cardinal coordinates to determine whether the planar surface is aligned with the cardinal coordinate system. If the planar surface is aligned with a cardinal direction (e.g., east, west, north, south) and the heading of the vehicle is known, the planar surface can still be used to correct the orientation sensor. If the change between the axis of the planar surface and the cardinal axis is greater than the threshold (indicating that the planar surface is not aligned with a cardinal direction), the planar surface is discarded at step 408. If the change between the axis of the planar surface and the cardinal axis is less than the threshold (indicating that the planar surface is aligned with a cardinal direction), the planar surface is used to correct the orientation sensor at step 416.

[0033] In some embodiments, after determining at step 414 or step 418 that the planar surface is a good candidate for correcting the orientation sensor, an orientation feedback value is generated to be provided in feedback to the orientation sensor. That is, the positive determination at step 414 or step 418 indicates that the detected surface is planar and that one or more axes associated with the detected surface area are aligned with one or more axes of the vehicle. As a result of this assumption, detected changes or discrepancies between the one or more axes defined by the planar surface and the orientation data provided by the orientation sensor are attributed to errors of the orientation sensor. In some embodiments, based on these detected errors, orientation feedback can be provided to correct the orientation data provided by the orientation sensor. As described above with respect to Figure 1 and Figure 3 In some embodiments, this can include providing the orientation feedback directly to the orientation sensor itself to reinitialize the orientation sensor or align the orientation sensor with the corrected orientation. In other embodiments, for example, as shown in Figure 3 the orientation feedback is latched into a correction unit, such as the IMU correction unit 308, which accumulates corrections provided in response to the detected changes.

[0034] In some embodiments, the provided orientation feedback is referred to as a “tweak” because it is used to slowly modify (i.e., tweak) the corrections applied to the orientation sensor to prevent large changes in the corrections applied to the orientation data. The magnitude of the orientation feedback can be determined in various ways. For example, in some embodiments, a fixed value is assigned regardless of the magnitude of the detected changes between the respective axes. The fixed value is typically small and ensures that each adjustment is relatively small in nature. In other embodiments, a variable tweak value is assigned based on the magnitude of the axis changes. Steps 414 and 418 ensure that the magnitude of the changes is less than a threshold value used in those steps, but this can still result in large adjustments to the orientation data. In some embodiments, step 416 is utilized to ensure that the magnitude of the orientation feedback does not exceed a threshold value. In some embodiments, the magnitude of the orientation feedback is based on the magnitude of the changes as well as a confidence level associated with the detected planar surface (determined at step 410). That is, for surfaces that are determined to be very flat and thus assigned a high confidence level, a larger magnitude of correction can be applied. Conversely, surfaces that are determined to be only somewhat flat and thus assigned a low confidence level will result in a lower magnitude correction being applied.

[0035] In other embodiments, as shown in Figure 4 a tradeoff between a fixed tweak value and a dynamic tweak value is provided. In Figure 4In the illustrated embodiment, at step 416, a change between one or more axes detected relative to the planar surface and one or more orientation axes defined by the orientation sensor is compared to a fine tuning threshold. If the change is less than the fine tuning threshold, which indicates a relatively small difference between the planar surface axes and the orientation axes, then at step 422 the magnitude of the change is used as orientation feedback provided to the IMU correction unit. If the change is greater than the fine tuning threshold, which indicates a large difference between the planar surface axes and the orientation axes, then at step 420 a fixed value is used as orientation feedback provided to the IMU correction unit. In some embodiments, the magnitude of the fixed value is equal to the magnitude of the fine tuning threshold. In this way, step 416 ensures that the fixed value utilized at step 420 represents the maximum orientation correction provided in the feedback to the IMU correction unit 308.

[0036] The process is repeated for each surface detected based on distance data provided by the one or more distance sensors. Orientation data measured by the one or more orientation sensors is corrected by accumulation of the orientation feedback values over time. In this way, corrections are provided to prevent errors from accumulating in the orientation sensors.

[0037] Discussion of Possible Embodiments

[0038] The following is a non-exclusive description of possible embodiments of the present invention.

[0039] According to some aspects, an orientation system mounted on a vehicle includes an orientation sensor, a distance sensor, and a vehicle processing unit. The orientation system is configured to generate orientation data, and the distance sensor is configured to generate relative distance data measuring a relative distance to an object outside the vehicle. The vehicle processing unit is configured to receive the orientation data from the orientation sensor and the relative distance data from the distance sensor, wherein the vehicle processing unit detects an orientation error based on the relative distance data.

[0040] The system of the preceding paragraph can optionally include (additionally and / or alternatively) any one or more of the following features, configurations and / or additional components.

[0041] For example, in some embodiments, the vehicle processing unit can provide orientation feedback to the orientation sensor to correct for the detected orientation error.

[0042] In some embodiments, the vehicle processing unit can be configured to detect a planar surface based on the received relative distance data, and to define one or more axes associated with the detected planar surface.

[0043] In some embodiments, the vehicle processing unit can further compare an orientation axis derived from the orientation data to the one or more axes associated with the detected planar surface.

[0044] In some embodiments, the vehicle processing unit can generate orientation feedback in response to the change between the compared axes being less than a threshold.

[0045] In some embodiments, the orientation feedback is assigned a magnitude equal to the change in response to the change being less than a fine-tune threshold.

[0046] In some embodiments, the orientation feedback is assigned a predetermined magnitude in response to the change being greater than the fine-tune threshold.

[0047] In some embodiments, the predetermined magnitude can be equal to the fine-tune threshold.

[0048] In some embodiments, the orientation feedback can be provided in feedback to the orientation sensor.

[0049] In some embodiments, the orientation feedback can be provided in feedback to an orientation correction unit that accumulates corrections over time.

[0050] In some embodiments, the distance sensor can include one or more of: a radar-based sensor, a LiDAR-based sensor, a laser-based sensor, and a vision-based sensor.

[0051] According to another aspect, there is provided a method of correcting orientation errors accumulated in an orientation sensor, the method comprising: receiving orientation data defining an orientation of a vehicle along one or more orientation axes; and receiving relative distance data from a distance sensor, wherein the relative distance data comprises a measured relative distance to an external object. The method further comprises: detecting planar surfaces in the received relative distance data and one or more surface axes associated with the detected planar surfaces, and calculating a change between the orientation axes and the planar axes. One or more of the orientation axes are corrected in response to the calculated change being less than a threshold.

[0052] The system of the preceding paragraph can optionally include (additionally and / or alternatively) any one or more of the following features, configurations and / or additional steps.

[0053] In some embodiments, the step of detecting planar surfaces can further comprise detecting a change in the planar surfaces and discarding detected planar surfaces having a change greater than a second threshold.

[0054] In some embodiments, the method can further comprise: assigning a confidence value to the detected planar surfaces.

[0055] In some embodiments, correcting one or more of the orientation axes can further comprise: generating an orientation feedback value having a magnitude based on the assigned confidence value.

[0056] In some embodiments, correcting the one or more orientation axes can include generating an orientation feedback having a magnitude based at least in part on the computed change.

[0057] In some embodiments, if the computed change is less than a fine-tune threshold, the orientation feedback can be assigned a value equal to the change.

[0058] In some embodiments, if the computed change is greater than the fine-tune threshold, the orientation feedback can be assigned a fixed value.

[0059] In some embodiments, the fixed value can be equal to the fine-tune threshold.

[0060] In some embodiments, the orientation feedback is added to a running orientation correction value, wherein the running orientation correction value represents a cumulative of received orientation feedback.

Claims

1. An orientation system installed on a vehicle, the orientation system comprising: An orientation sensor configured to generate orientation data; A distance sensor configured to generate relative distance data that measures the relative distance to objects outside the vehicle; as well as A vehicle processing unit configured to receive orientation data from the orientation sensor and relative distance data from the distance sensor, wherein the vehicle processing unit detects an orientation error based on the relative distance data, and wherein the vehicle processing unit is configured to: Detecting planar surfaces based on received relative distance data. Define one or more axes associated with the detected planar surface. The orientation axis derived from the orientation data is compared with one or more axes associated with the detected planar surface. Orientation feedback is generated in response to changes in the compared axes being less than a threshold.

2. The orientation system as described in claim 1, characterized in that, The vehicle processing unit provides orientation feedback to the orientation sensor to correct detected orientation errors.

3. The orientation system as described in claim 1, characterized in that, In response to the change being less than the fine-tuning threshold, the orientation feedback is assigned an amount equal to the magnitude of the change.

4. The orientation system as described in claim 1, characterized in that, In response to the change being greater than the fine-tuning threshold, the orientation feedback is assigned a predetermined amplitude.

5. The orientation system as described in claim 4, characterized in that, The predetermined range is equal to the fine-tuning threshold.

6. The orientation system as claimed in claim 1, characterized in that, The orientation feedback is provided in the feedback given to the orientation sensor.

7. The orientation system as claimed in claim 1, characterized in that, The orientation feedback is provided in the feedback to the orientation correction unit, which is corrected cumulatively over time.

8. The orientation system as claimed in claim 1, characterized in that, The distance sensor includes one or more of the following: radar-based sensor, LiDAR-based sensor, laser-based sensor, and vision-based sensor.

9. A method for correcting orientation errors accumulated in an orientation sensor, the method comprising: Receive orientation data, which defines the orientation of the vehicle along one or more orientation axes; Receive relative distance data from a distance sensor, wherein the relative distance data includes measured relative distances to external objects; Detect planar surfaces in the received relative distance data and one or more surface axes associated with the detected planar surfaces; Calculate the change between the orientation axis and the planar axis; as well as If the calculated change is less than a threshold, then one or more of the orientation axes are corrected.

10. The method as described in claim 9, characterized in that, Detecting a planar surface includes: detecting changes in the planar surface and discarding detected planar surfaces with changes greater than a second threshold.

11. The method as described in claim 10, characterized in that, Further includes: Assign confidence values ​​to the detected planar surfaces.

12. The method as described in claim 11, characterized in that, Correcting one or more orientation axes includes generating an orientation feedback value with an amplitude based on the assigned confidence value.

13. The method as described in claim 9, characterized in that, Correcting one or more orientation axes includes generating an orientation feedback with a magnitude based at least in part on the calculated changes.

14. The method as described in claim 13, characterized in that, If the calculated change is less than the fine-tuning threshold, the orientation feedback is assigned a value equal to the change.

15. The method as described in claim 14, characterized in that, If the calculated change is greater than the fine-tuning threshold, the orientation feedback is assigned a fixed value.

16. The method as described in claim 15, characterized in that, The fixed value is equal to the fine-tuning threshold.

17. The method as described in claim 13, characterized in that, The orientation feedback is added to the running orientation correction value, where the running orientation correction value represents the accumulation of the received orientation feedback.

Images