Sensor pod assembly

Abstract

A sensor pod system includes one or more sensor pods with multiple sensors configured to collect data from an environment. The sensor pods may have effective fields of view created by individual sensors with overlapping fields of view. The sensor pod system may include sensors of different types and modalities. The sensor pods of the sensor pod system may be modularly installed on a vehicle, e.g., an autonomous vehicle, to collect and provide data of the environment while the vehicle is operating.

Description

Technical Field

[0001] The present invention relates to a sensor pod assembly.

Background Art

[0002] Related Applications This application claims priority to U.S. Patent Application Nos. 16 / 864,082, 16 / 864,109, 16 / 864,122, 16 / 864,138, and 16 / 864,146, each filed on April 30, 2020, which are hereby incorporated by reference herein.

[0003] Many currently operating vehicles are designed to use sensors to perceive their surroundings. Sensors are often integrated into the vehicle, for example, within the vehicle's body panels. However, integration into the vehicle body often limits the sensor's field of view. In other examples, sensors can be attached outside the vehicle, such as on the vehicle's roof. However, installing sensors outside the vehicle increases the potential for the sensors to collide with external objects, potentially damaging the sensors and / or the collided objects. These and other problems are compounded by the number and types of sensors to be included in the vehicle. Although sensor technology is improving, small, electric, two-way, and / or autonomous vehicles can have unique components and configurations where conventional sensor systems may be insufficient to provide data to the vehicle during operation or where long delays can occur while sensors are being replaced.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Patent Document 3

[0005] [Non-Patent Document 1] Regulation (EC) No. 78 / 2009 of the European Parliament and the Council of 14 January 2009 [Non-Patent Document 2] European New Vehicle Assessment Programme Pedestrian Test Protocol, Version 8.4, November 2017 [Brief explanation of the drawing]

[0006] A detailed explanation will be provided with reference to the attached diagram. In the diagram, the leftmost digit of the reference number identifies the diagram in which the reference number first appears. The use of the same reference number in different diagrams indicates similar or identical components or features. [Figure 1] This is a diagram illustrating an exemplary vehicle having one or more sensor pods that collect data and provide it to an autonomous vehicle, as illustrated by the present disclosure. [Figure 2] This is a top view of an exemplary vehicle having a sensor pod, as described in the present disclosure. [Figure 3] This is a simplified diagram of Figure 2 showing the field of view and optical axis of the sensor pod. [Figure 4] Figures 1 to 3 are end views of an exemplary vehicle. [Figure 5] This is a top view of an exemplary sensor pod according to the examples of the present disclosure. [Figure 6] This is a perspective view of a sensor pod detached from its mount in an exemplary vehicle, as shown in the example of the present disclosure. [Figure 7] This is an exploded view of an exemplary sensor pod according to the examples of the present disclosure. [Figure 8]This is a perspective view of a sensor pod with an external housing removed to show internal components, as illustrated in the example of the present disclosure. [Figure 9] This is a perspective view of a sensor pod with some internal components, trim components, and a portion of the external housing removed to show the pedestrian protection system as illustrated in the example of the present disclosure. [Figure 10] Figures 9 and 10 show schematic diagrams of energy management processes and techniques, based on the influence of energy management structures in the context of the sensor pod. [Figure 11] This diagram illustrates the process and technique for calibrating a sensor pod, as illustrated by the present disclosure. [Figure 12] This diagram illustrates the process and technique for calibrating a sensor pod, as illustrated by the present disclosure. [Figure 13] This diagram illustrates the process and technique for calibrating a sensor pod, as illustrated by the present disclosure. [Figure 14] These are block diagrams of exemplary systems for implementing the techniques shown in Figures 11 to 13. [Figure 15] This is an enlarged perspective view of the area shown in Figure 8. [Figure 16] This diagram illustrates a process and technique for cleaning the sensors of a sensor pod, as illustrated by the present disclosure. [Modes for carrying out the invention]

[0007] As described above, conventional sensor installations integrated into the vehicle body may not provide sufficient sensor coverage and replacement can be time-consuming. For example, in the case of sensors embedded in the body, such as the body panel, the process of removing and replacing the sensor often requires the removal of the vehicle's body panel and / or other parts. This can be a relatively long and complex process that may prevent the vehicle from being used. When sensors are mounted externally to the vehicle, the sensor may collide with an external object, potentially damaging the sensor and / or the object it hits.

[0008] This application relates to structures and techniques for improving the installation, packaging, maintenance, and replacement of sensors while providing protection to pedestrians. In an example, the sensor may be disposed within a self - contained assembly or “sensor pod” removably coupled to a vehicle. A variety of sensor pods, for example, four, may be disposed around the exterior of a vehicle to provide coverage to the environment surrounding the vehicle.

[0009] For example, a sensor pod may include a frame having an attachment interface for removably coupling the sensor pod to the vehicle. In an example, the sensor pod may include a variety of sensors mounted at positions on the frame, where each position provides a field of view to each sensor that complements the fields of view of other sensors in the sensor pod to create an effective field of view for each sensor pod. In an example, the sensor pod may have a variety of sensor types. For example, some of the sensors may be imaging sensors, such as cameras (e.g., RGB cameras, monochrome cameras, intensity (grayscale) cameras, infrared cameras, ultraviolet cameras, depth cameras, stereo cameras, time - of - flight (TOF) sensors, etc.), while other sensors may be ranging or distance sensors, such as one or more ultrasonic transducers of a light detection and ranging (LIDAR) sensor, a radio detection and ranging (RADAR) sensor, a sound navigation and ranging (SONAR) sensor, or another known sensor type. Other types of sensors, such as inertial measurement sensors, may be included in the sensor pod as an addition or alternative. In an example, sensors of the same type within the sensor pod may have different and / or overlapping fields of view to provide coverage to a portion of the environment surrounding the vehicle.

[0010] In an example, the frame is a casting and can provide sensor locations through the casting surface. In an example, the casting frame provides a rigid mount to the sensor and spaces the sensor within the pod slightly away from the vehicle. The casting surface can provide a mounting interface with sufficient accuracy without requiring a secondary process to machine the surface. The use of such a mounting surface can reduce the computational resources required for calibration in some examples (e.g., by ensuring that the sensor is placed within some known tolerance) and by ensuring little movement when operating the vehicle, thereby reducing the number of calibrations required.

[0011] In an example, the sensor pods can have a cleaning system with nozzles disposed adjacent to them for cleaning the sensor surfaces. In an example, the nozzles can apply pressurized fluid for cleaning the surface. In an example, the cleaning system can be supplied by a centralized fluid reservoir that can supply various sensor pods. The fluid provided to a particular sensor pod can be distributed to various sensors within the sensor pod via a fluid manifold.

[0012] In an example, the sensor pods have a supply harness that can connect the sensor pods to the vehicle. The supply harness can provide power and pressurized fluid to the sensor pods. In an example, the sensor pods can also have a sensor harness that is electrically coupled to the vehicle for transmitting sensor data from the sensors of the sensor pods to a computing system of the vehicle and is electrically coupled to the sensors.

[0013] As described, in the example, the vehicle may have a variety of sensor pods mounted on the vehicle body. For example, the vehicle may include a body having a first end along the longitudinal axis and a second end distal to the first end, with the first sensor pod detachably coupled to the body at a first position. In the example, the first position is at an altitude above the ground interface, adjacent to the first end, and spaced apart from the longitudinal axis in a first direction along the first transverse axis. In the example, the vehicle may include second, third, and fourth sensor pods distributed around the vehicle body (for example, adjacent to the four corners or quarter sections of the vehicle). In the example, the first, second, third, and fourth sensor pods each have an effective field of view. When mounted on the vehicle, the effective field of view of a sensor pod overlaps with the effective field of view of an adjacent sensor pod. In the example, overlapping fields of view allow the sensor pods to see around the entire vehicle. In the example, overlapping fields of view provide redundancy to the sensors. For example, if a sensor pod is damaged or fails, the other three sensor pods still provide an effective field of view to see around the entire vehicle. In the example, the effective field of view of the sensor pods is at least 270 degrees. In the example, the elevation of the sensor pods can be high enough to avoid or see over most obstacles commonly encountered while driving. In the example, the sensor pods can be mounted at an elevation of at least about 4 feet and no more than about 7 feet above the ground. In some examples, the sensor pods can be mounted close to or slightly below the top of the vehicle. For example, the mount for each sensor pod can be coupled to the vehicle within about 18 inches of the top of the vehicle, and the sensor pods can extend vertically above and / or below the mounting position. In one particular example, the sensor pods can be mounted on the vehicle at about 5 feet 10 inches above the ground. In the example, the sensor pods can extend above the top of the vehicle. In the example, the sensor pod may be mounted on top of the vehicle. In the example, the sensor pod is mounted at least about 5 feet 10 inches above the ground, but less than 1 foot above the top of the vehicle body.

[0014] In the example, the sensor pod may include a pedestrian protection system. For example, the sensor pod may include a frame coupled to a vehicle, sensors coupled to the frame, and a collision structure coupled to the frame. In the example, the collision structure comprises an outer surface and a collision energy absorption structure. The outer surface is configured to interface with pedestrians or other objects during a collision. At least a portion of the outer surface is positioned outside the sensor relative to the frame. The collision energy absorption structure is positioned between the outer surface and the frame and is configured to absorb a portion of the energy transmitted from the collision through the outer surface. In the example, the collision energy absorption structure includes a stress concentration zone, where the stress concentration zone causes local plastic deformation of the collision energy absorption structure to absorb energy from the collision. In the example, the outer surface may be configured to move substantially uniformly during a collision. In the example, the outer surface may be configured to deform upward during a collision to absorb some of the energy from the collision. In the example, the sensor pod may have another sensor positioned on top of the frame and coupled to the frame by a deformable fastener. In the example, a deformable fastener can deform during an impact to absorb energy from the impact. In the example, if the energy from the impact exceeds a threshold, the deformable fastener can release the sensor. In the example, the sensor pod may include a baffle structure, which includes a crumple zone configured to plastically deform during impact to absorb energy.

[0015] While several examples are given in the context of a vehicle having sensor pods positioned close to the four corners of the vehicle and at an altitude close to the top surface of the vehicle, other examples may use other numbers and configurations of sensor pods, and / or sensor pods may be positioned at other locations relative to the vehicle (e.g., altitude, lateral spacing, and / or longitudinal spacing). Furthermore, while the exemplary sensor pods provided herein include a particular combination of various different types of sensors, other examples may incorporate a smaller number of sensors of some types and additional sensors of other types. Other examples may include several more sensors positioned on the sensor pod to provide a desired resolution or redundancy.

[0016] Exemplary autonomous vehicle with sensor pod Figure 1 is a diagram of an exemplary vehicle 100 having one or more sensor pod assemblies composed of various sensors for collecting information about the surroundings of the autonomous vehicle, according to an example of the present disclosure. The vehicle shown in Figure 1 may be a bidirectional autonomous vehicle configured to operate according to the Level 5 classification issued by the U.S. Department of Transportation's National Highway Traffic Safety Administration, which describes a vehicle capable of performing all safety-critical functions throughout the entire journey in a state in which it is not expected that a driver (or occupant) will control the vehicle at any given time. However, in other examples, the vehicle may be a fully or partially autonomous vehicle having any other level or classification. Furthermore, in some examples, the energy management structures described herein may also be applicable to non-autonomous and / or non-bidirectional vehicles. Also, although examples are given in which the vehicle is a land vehicle, the techniques described herein are also applicable to aerial, marine, and other vehicles.

[0017] In the illustrated example, vehicle 100 includes a first sensor pod assembly 102A and a second sensor pod assembly 102B (collectively, “Sensor Pod 102”) coupled to a main body 104. Each of the sensor pod assemblies 102 in this example may include a variety of sensors and a system or structure for cleaning the sensor pod or other objects or protecting them during a collision. For example, sensor pod assembly 102A includes a first sensor 106, a second sensor 108, and a third sensor 110. In some examples, these sensors may be of a first type, such as an imaging sensor. In some examples, sensor pod assembly 102A also includes a fourth sensor 112 and a fifth sensor 114. In some examples, these sensors may be of a second type, such as a LIDAR (light detection and ranging) sensor.

[0018] In some examples, the sensor pod 102A also has an outer shell 116 or trim. In some examples, the outer shell 116 incorporates an energy-absorbing structure that can mitigate damage to objects colliding with the sensor pod assembly 102A.

[0019] In some examples, the main body 104 has an additional sensor 118 that is disposed on the main body 104, separate from the sensor pod assembly 102.

[0020] In some examples, the vehicle 100 may have a longitudinal axis 120 that spans the length of the vehicle 100. In some examples, the sensor pod assembly 102A may be positioned on the main body 104 at an altitude 122 above the ground interface 124. In some examples, the sensor pod assembly 102A may be positioned at a height or altitude above the vehicle such that the sensors of the sensor pod assembly can see over most obstacles that the vehicle may encounter. For example, seeing over obstacles may be advantageous, by which the vehicle's system can not only observe or react to an obstacle, but also to any additional obstacles or conditions beyond that obstacle. In some examples, observing an obstacle from a high angle may be advantageous, where the high angle may allow the vehicle's system to better determine one or more of the following, in particular, distance, approach speed, status, and direction.

[0021] In one example, the sensor pod 102A may be mounted at an altitude of at least approximately 4 feet above the ground, and no higher than approximately 7 feet. In another example, the sensor pod 102A may be mounted close to or slightly below the top of the vehicle. For example, the sensor pod mount may be coupled to the vehicle within approximately 18 inches above the top of the vehicle, and the sensor pod 102A may extend vertically above and / or below the mounting position. In one particular example, the sensor pod 102A is mounted on the vehicle at approximately 5 feet 10 inches above the ground. In another example, the sensor pod 102A extends above the top of the vehicle. In yet another example, the sensor pod 102A is mounted on top of the vehicle. In yet another example, the sensor pod 102A is mounted at least approximately 5 feet 10 inches above the ground, but less than 1 foot above the top of the vehicle body.

[0022] In some cases, the sensor pod can be mounted higher than 5 feet above the ground. In some cases, the sensor pod can be mounted higher than 5 feet 6 inches above the ground. In some cases, the sensor pod can be mounted higher than 5 feet 10 inches above the ground. In some cases, the sensor pod can be mounted higher than 6 feet above the ground. In some cases, the sensor pod can be mounted higher than 6 feet 6 inches above the ground. In some cases, the sensor pod can be mounted higher than 7 feet above the ground.

[0023] In some cases, the sensor pod assembly is mounted below an altitude threshold. In some cases, it may be beneficial to raise the sensor, and it may also be beneficial to mount the sensor pod assembly below a certain altitude. For example, if mounted too high, the sensor pod assembly may be at risk of colliding with elevated obstacles, particularly trees, signs, bridges, overpasses, power lines, and wiring. In some cases, if mounted too high, the sensor pod assembly may require additional support to give it sufficient rigidity or stiffness to reduce shaking or amplified vibrations. In some cases, the sensor pod may be mounted lower than 2 feet above the roof of the vehicle. In some cases, the sensor pod may be mounted lower than 18 inches above the roof of the vehicle. In some cases, the sensor pod may be mounted lower than 1 foot above the roof of the vehicle. In some cases, the sensor pod may be mounted lower than 6 inches above the roof of the vehicle. In some cases, the sensor pod may be mounted so that it is substantially washed off the roof of the vehicle. In some cases, the sensor pod may be mounted below the roof of the vehicle.

[0024] In this example, the vehicle 100 includes one or more computer systems 126 to control the operation of one or more systems of the vehicle 100. For example, in the case of an autonomous vehicle, the computer system 126 may include one or more processors and memory and may be configured in particular to control the vehicle 100 to receive and process sensor data from one or more sensors and to plan a route for the vehicle through the environment.

[0025] In some examples, the computer system 126 controls the operation of one or more systems of the vehicle 100. For example, in the case of an autonomous vehicle, the computer system 126 may include one or more processors 128 and a memory 130 communicably coupled to one or more processors 128, and may be configured in particular to control the vehicle 100 to receive and process sensor data from one or more sensors and to plan a route for the vehicle through the environment. In some examples, the computer system 126 may also include a controller 132 configured to control subsystems of the vehicle 100. For example, the controller 132 may control a sensor cleaning system 134. In some examples, the cleaning system 134 may comprise a reservoir, fluid, pressurizer, pump, and valve, and may be connected to one or more of the sensor pods 102. In the example, the vehicle 100 has a variety of cleaning systems 134. In the example, each of the cleaning systems 134 is coupled to one or more of the sensor pods 102. In the example, one of the various cleaning systems 134 is coupled to a first subset of sensor pods 102, for example, sensor pod 102A, and another of the various cleaning systems 134 is coupled to a second subset of sensor pods 102, for example, sensor pod 102B. In the example, the first and second subsets of sensor pods 102 are distinct. For example, a cleaning system 134 coupled to sensor pod 102A does not directly provide cleaning functionality to sensor pod 102B. In the example, the first and second subsets of sensor pods 102 partially overlap. In the example, the first and second subsets of sensor pods 102 completely overlap. For example, the various cleaning systems 134 provide redundancy to each other. In some examples, the cleaning system 134 is located within the main body 104. In some examples, the cleaning system 134 is located within a detachable part of the vehicle 100, for example, within a drive module.

[0026] In the illustrated example, vehicle 100 is an autonomous vehicle, but vehicle 100 could be any other type of vehicle, such as a semi-autonomous vehicle, or any other system having at least an image capture device (e.g., a smartphone with a camera). Although shown in Figure 1 as residing in the main body 104 for illustrative purposes, it is intended that the computer system 126 is accessible to vehicle 100 (for example, stored in memory away from vehicle 100, such as on the memory of a remote computer device, or otherwise accessible). In some examples, a variety of computer systems 126 may be included on vehicle 100. In some examples, the computer system 126 may be located within the main body 104, the drive assembly, or a combination thereof.

[0027] The processor 128 of the vehicle 100 may be any suitable processor capable of processing data and executing instructions to perform operations as described herein. For example, but not limited to, the processor 128 may comprise one or more central processing units (CPUs), graphics processing units (GPUs), or any other devices or parts of devices that process the electronic data for conversion into other electronic data that can be stored in registers and / or memory. In some examples, integrated circuits (e.g., ASICs), gate arrays (e.g., FPGAs), and other hardware devices may also be considered processors, insofar as they are configured to implement encoded instructions.

[0028] Memory 130 is an example of a non-temporary computer-readable medium. Memory 130 may store an operating system and one or more software applications, instructions, programs, and / or data to implement the methods and functions resulting from various systems described herein. In various implementations, memory may be implemented using any suitable memory technology, such as static random-access memory (SRAM), synchronous dynamic RAM (SDRAM), non-volatile / flash-type memory, or any other type of memory capable of storing information. The architectures, systems, and individual elements described herein may include many other logical, programmatic, and physical components, and those shown in the accompanying figures are merely examples relevant to the description herein.

[0029] In some examples, memory 130 may include at least working memory and storage memory. For example, working memory may be a limited-capacity, high-speed memory (e.g., cache memory) used to store data to be operated on by processor 128. In some cases, memory 130 may include storage memory, which may be a relatively large-capacity, lower-speed memory used for long-term storage of data. In some cases, processor 128 may not be able to operate directly on data stored in storage memory as described herein, and the data may need to be loaded into working memory for operations to be performed on the data.

[0030] Figure 2 shows the vehicle 100 of Figure 1 from a top-down view. In the illustrated example, the vehicle 100 includes a first sensor pod assembly 102A1, a second sensor pod assembly 102B2, a third sensor pod assembly 102A3, and a fourth sensor pod assembly 102B4 (collectively, “Sensor Pod 102”) coupled to the main body 104. Figure 2 also shows a sensor 118 disposed on the vehicle away from the sensor pod 102. Figure 2 shows a longitudinal axis 120 along the length of the vehicle 100, where the vehicle has a first end 200 along the longitudinal axis 120 and a second end 202 distal to the first end. Figure 2 shows a first transverse axis 204 at the first end 200 and a second transverse axis 206 at the second end 202.

[0031] Figure 2 shows a first sensor pod assembly 102A1, which is removablely mounted to the vehicle 100 at a first position 208, located adjacent to the first end 200, and spaced away from the longitudinal axis in a first direction 210 along the first transverse axis 204. Figure 2 shows a second sensor pod assembly 102B2, which is removablely mounted to the vehicle at a second position 212, located adjacent to the first end 200, and spaced away from the longitudinal axis 120 in a second direction 214 opposite to the first direction 210 along the first transverse axis 204. Figure 2 shows a third sensor pod assembly 102A3, which is removablely mounted to the vehicle at a third position 216, located adjacent to the second end 202, and spaced away from the longitudinal axis 120 in a first direction 210 along the second transverse axis 206. Figure 2 shows a fourth sensor pod assembly 102B4 that is removablely mounted to the vehicle at a fourth position 218, located adjacent to the second end 202, and spaced apart from the longitudinal axis 120 in a first direction 210 along the second transverse axis 206.

[0032] In some examples, the sensor pods 102 may be spaced apart from each other along the first or second transverse axis by a distance of 50, 60, or 70 inches. In some examples, the sensor pods 102 may be spaced apart from each other along the first or second transverse axis and remain within 2 inches outside the body 104, within 5 inches outside the body 104, within 8 inches outside the body 104, within 10 inches outside the body 104, or within 12 inches outside the body 104.

[0033] In some examples, the sensor pods 102 may be spaced apart from each other along the vertical axis 120 by a distance of 100, 110, or 120 inches. In some examples, the sensor pods 102 may be spaced apart from each other along the vertical axis 120 and remain within 2 inches outside the main body 104, within 5 inches outside the main body 104, within 8 inches outside the main body 104, within 10 inches outside the main body 104, or within 12 inches outside the main body 104.

[0034] In some examples, the sensor pod 102 has a variety of sensors, each with its own sensor field of view. In some examples, the fields of view of the individual sensors can be combined to create an effective sensor field of view for the sensor pod. Figure 2 shows exemplary sensor fields of view of the individual sensors of the sensor pod 102. For example, the sensor pod assembly 102B2 has a first sensor field of view 224 with optical axis 226, a second sensor field of view 228 with optical axis 230, and a third sensor field of view 232 with optical axis 234. In the example, these sensor fields of view can be combined to create an effective field of view 236 for the sensor pod assembly. In the example, the first sensor field of view 224 is approximately 80 degrees. In the example, the first sensor field of view 224 is between 75 and 85 degrees. In the example, the first sensor field of view 224 is between 60 and 90 degrees. In the example, the second sensor field of view 228 is approximately 80 degrees. In the example, the field of view of the second sensor 228 is between 75 and 85 degrees. In the example, the field of view of the second sensor 228 is between 60 and 90 degrees.

[0035] In the example, the field of view 232 of the third sensor is approximately 60 degrees. In the example, the field of view 232 of the third sensor is between 60 and 65 degrees. In the example, the field of view 232 of the third sensor is between 50 and 90 degrees.

[0036] In the example, the optical axis 226 is substantially parallel to the vertical axis 120. In the example, the optical axis 226 is angled away from the vertical axis 120. For example, the optical axis 226 is angled about 5 degrees outward from the vertical axis 120. In the example, the optical axis 226 is angled between 2 and 10 degrees outward from the vertical axis 120. In the example, the optical axis 226 is angled about 5 degrees inward from the vertical axis 120. In the example, the optical axis 226 is angled between 2 and 10 degrees inward from the vertical axis 120. In the example, if the optical axis 226 is angled inward or outward from the vertical axis 120, the chances of both sensors in two adjacent sensor pods pointed towards the same end of the vehicle experiencing sun blindness are greatly reduced. For example, the lens surfaces of both sensors will not be parallel and will not receive sunlight at the same angle.

[0037] Figure 2 also shows a sensor 118 having a sensor field of view. In the example, sensor 118 may be similar to other sensors 118. In the example, sensor 118 may have different characteristics and / or features. For example, sensor 118 may be sensor 238 and have a sensor field of view 240 with optical axis 242. In the example, sensor 118 may be sensor 244 and have a sensor field of view 246 with optical axis 248. In the example, sensor field of view 240 may be larger than sensor field of view 246. For example, sensor 244 may give a relatively narrow field of view compared to sensor 238. In the example, sensor 244 may give more information about positions further away from vehicle 100 along optical axis 248. As a practical example, sensor 244 may give information about conditions further ahead in the direction of travel. In the example, sensor field of view 240 is between 140 and 170 degrees. In the example, sensor field of view 246 is about 30 degrees.

[0038] In some examples, the sensor pod 102 may be spaced away from the outer edge of the vehicle 100. For example, Figure 2 shows a sensor pod 102A1 spaced 250 away from the main body 104, e.g., a fender, grille, or other outer surface. In this example, the distance 250 is approximately 285 mm. In this example, the distance 250 is between 200 mm and 300 mm. In this example, the distance 250 is between 50 mm and 200 mm. In this example, the distance 250 is between 0 mm and 50 mm. In this example, the distance 250 is negative. For example, the distance 250 is between -1 mm and -100 mm. In this example, the surface of the sensor pod will extend beyond the main body 104.

[0039] Figure 2 also shows the sensor pod 102A1 spaced 252 away from most of the side of the main body 104, for example, a fender, door, or other outer surface. In this example, the distance 252 is approximately 10 mm. In this example, the distance 252 is between 0 mm and 10 mm. In this example, the distance 252 is negative. For example, the distance 252 is between -1 mm and -20 mm. In this example, most of the surface of the side of the main body 104 will extend beyond the sensor pod. In this example, the distance 252 is approximately -12 mm.

[0040] Figure 3 shows the vehicle 100 of Figures 1 and 2. In the illustrated example, the vehicle 100 includes a first sensor pod assembly 102A1, a second sensor pod assembly 102B2, a third sensor pod assembly 102A3, and a fourth sensor pod assembly 102B4 (collectively, “Sensor Pod 102”) coupled to the main body 104. Figure 3 also shows the optical axis of the sensors of the sensor pods and the effective field of view 236 of the sensor pods for each of the sensor pods 102. In the example, the sensor pod 102 has additional sensors. For example, the sensor pod 102 may have a LIDAR sensor with a field of view 300 of a second sensor type. In the example, the field of view of the second sensor type may cover the same or substantially similar field of view.

[0041] In some cases, the effective sensor field of view 236 of the first sensor pod assembly 102A1 overlaps with at least a portion of the effective sensor field of view 236 of the second sensor pod assembly 102B2 and at least a portion of the effective sensor field of view 236 of the fourth sensor pod assembly 102B4. In some cases, the effective sensor field of view 236 of the second sensor pod assembly 102B2 overlaps with at least a portion of the effective sensor field of view 236 of the first sensor pod assembly 102A1 and at least a portion of the effective sensor field of view 236 of the third sensor pod assembly 102A3. In some cases, the effective sensor field of view 236 of the third sensor pod assembly 102A3 overlaps with at least a portion of the effective sensor field of view 236 of the second sensor pod assembly 102B2 and at least a portion of the effective sensor field of view 236 of the fourth sensor pod assembly 102B4. In some examples, the effective sensor field of view 236 of the fourth sensor pod assembly 102B4 overlaps with at least a portion of the effective sensor field of view 236 of the first sensor pod 102A1 and overlaps with at least a portion of the effective sensor field of view 236 of the third sensor pod assembly 102A3.

[0042] In the example, the effective sensor pod field of view can be greater than 270 degrees. In the example, the effective sensor fields of view of adjacent sensor pod assemblies overlap at a certain distance from the vehicle. For example, the distance could be 5 feet, 2 feet, or 1 foot. In the example, a distance of 0 means that the sensor fields of view overlap within the vehicle, and a negative distance means that the sensor fields of view overlap on the vehicle. In the example, this creates a 360-degree effective system field of view for the sensor pods, enabling coverage around the vehicle. In the example, the overlap is such that the effective system field of view of the sensor pods is 360 degrees even when 3 out of 4 sensor pods are active. In the example, the overlap is such that the effective system field of view of the sensor pods is 360 degrees even when 2 out of 4 sensor pods are active, if the two active sensor pods are at opposite corners of the vehicle or in the opposite quarter.

[0043] In the example, the sensor pod 102 extends from the main body 104. For example, each sensor pod protrudes at least a first distance from the longitudinal end of the main body 104 of the vehicle 100 and at least a second distance from the side. In the example, the first and second distances may be the same or different. In the example, the first and second distances may be 0 inches, 2 inches, 6 inches, 8 inches, 10 inches, 12 inches or more.

[0044] In the example, the sensor pod 102 is positioned in different quarter-sections of the vehicle, for example, near a corner. By being near the end of the vehicle, the sensor pod may have an advantageous field of view in some situations. For example, when turning a corner or exiting a parking lot, the sensor pod may be able to look around the corner to collect data before the entire vehicle enters the space. In the example, the sensor pod 102 is positioned at a first, second, third, and fourth position on the vehicle. For example, the first position is within a first distance from the first end, and the distance from the longitudinal axis of the first position is greater than the second distance from the longitudinal axis such that the effective sensor field of view of the first sensor pod includes a view of the vehicle and of an object behind an obstacle located at the first position away from the vehicle.

[0045] In the example, the sensor field of view 300 may be limited to less than 360 degrees for each individual sensor. For example, the sensor field of view may be greater than 270 degrees, approximately 270 degrees, or less than 270 degrees. In the example, the location of the field of view for some types of sensors may be beneficial for data acquisition and computation efficiency. For example, by placing sensors near the corners of the vehicle, a 270-degree field of view still maintains redundancy from overlapping fields of view while enabling a full 360-degree view when combined with other sensors. However, by limiting the sensor field of view, unnecessary data from immutable parts of the vehicle is not collected or is efficiently removed from the data early in processing, reducing the load on the system. In the example, it may be beneficial to view parts of the vehicle, and the field of view may be limited to below a vehicle-containing threshold.

[0046] In this example, the sensor pod 102 has two configurations or types. For example, a first type of sensor pod may be a mirror image of a second type of sensor pod. Figure 3 shows an example where sensor pod assemblies 102A1 and 102A3 are of the first type, while sensor pod assemblies 102B2 and 102B4 are of the second type.

[0047] Figure 4 shows the vehicle 100 from Figures 1 to 3 from an end view. In the illustrated example, the vehicle 100 includes a first sensor pod assembly 102A1. For simplicity, the description of the sensor's field of view refers to the coverage in a plan view using the vertical coverage over the field of view that is inherently included. Figure 4 shows an example in which the sensors of sensor pod 102 may be oriented within the sensor pod to correct the vertical orientation. In the example, the first sensor 400 and the second sensor 402 of sensor pod assembly 102A1 have a field of view and a detection axis. For example, the first sensor 400 may have a field of view 404 and a detection axis 406, and the sensor 402 may have a field of view 408 and a detection axis 410. In this example, the detection axis 406 may be substantially horizontal with respect to the vehicle 100, while the detection axis 410 may be tilted downwards, for example, at 45 degrees. In the example, the field of view 404 of the first sensor 400 may provide a view of objects further away from the vehicle compared to the field of view 408 of the second sensor 402, which may provide a view of objects closer to the vehicle. In the example, the second sensor 402 may provide useful information when the vehicle is operating near an object, for example, when parked near a person or object or operating near one.

[0048] Figure 4 also shows a second sensor pod assembly 102B2. In this example, sensor pod 102B2 is similar to sensor pod 102A1. In this example, sensor pod 102B2 has a mirrored layout compared to sensor pod 102A1. In this example, sensor pod 102B2 has a height of 412 and a width of 414. In this example, the height 412 may be approximately 330 mm. In this example, the height 412 may be between 300 mm and 350 mm. In this example, the height 412 may be between 275 mm and 375 mm.

[0049] In the example, a width of 414 could be approximately 230 mm. In the example, a width of 414 could be between 200 mm and 250 mm. In the example, a width of 414 could be between 175 mm and 175 mm.

[0050] Exemplary sensor pod assembly Figure 5 shows a top-down view of an exemplary sensor pod 500. In this example, the sensor pod 500 includes a first sensor 502, a second sensor 504, and a third sensor 506. In some examples, these sensors may be of a first type, such as an imaging sensor. In some examples, the sensor pod 500 also includes a fourth sensor 508 and a fifth sensor 510 (below). In some examples, these sensors may be of a second type, such as a LiDAR (light detection and ranging) sensor.

[0051] In the example, the sensor pod 500 also has an outer shell 514 or trim. In the example, the outer shell 514 incorporates an energy-absorbing structure that can mitigate damage to an object colliding with the sensor pod 500.

[0052] In this example, the sensor pod 500 is similar to the sensor pod 102B2 shown in Figure 4. In this example, the sensor pod 500 has a width of 516 and a length of 518. In this example, the width 516 is similar to the width 414 shown in Figure 4. In this example, the width 516 may be approximately 230 mm. In this example, the width 516 may be between 200 mm and 250 mm. In this example, the width 516 may be between 175 mm and 175 mm.

[0053] In the example, length 518 could be approximately 300 mm. In the example, length 518 could be between 250 mm and 350 mm. In the example, length 518 could be between 225 mm and 375 mm.

[0054] In the example, there is a width of 412 and a height of 414. In the example, the height of 412 could be approximately 330 mm. In the example, the height of 412 could be between 300 mm and 350 mm. In the example, the height of 412 could be between 275 mm and 375 mm.

[0055] Figure 6 shows a perspective view of an exemplary example of a vehicle 600 with a sensor pod 602 mounted on the vehicle 600. In the example, the sensor pod 602 may be mounted on the vehicle 600 through a mounting system 604. In the example, a mounting arm 606 may be coupled to the vehicle in a rigid and permanent configuration. The sensor pod 602 may be mounted on the mounting arm 606 through a mounting interface 608 on the sensor pod 602. In the example, the mounting arm and mounting interface 608 have indexing features 610 to increase the accuracy and consistency of the orientation of the sensor pod 602 and the sensor relative to the vehicle 600 when installing and removing the sensor pod 602.

[0056] Figure 7 is an exploded view 700 of the sensor pod 602. In this example, the sensor pod 602 includes a frame 702 with various sensors, such as a first camera 704, a second camera 706, a third camera 708, a first LiDAR 710, and a second LiDAR 712. The sensor pod 602 may also include an exterior trim 714, a bracket 716, a paddle lamp 718, a seal 720, a cleaning system 722, and a harness 724.

[0057] The exploded view 700 also shows the pod arm 726. In this example, the pod arm 726 will be connected to a vehicle to which the sensor pod 602 may be later mounted.

[0058] Figure 8 shows a sensor pod 800 without an outer shell or trim. In this example, the sensor pod 800 includes a frame 802 having a mounting interface (not shown) that allows the sensor pod 800 to be mounted on a vehicle (not shown). In this example, the sensor pod 800 includes a first sensor 804 that is removablely mounted at a first position 806 of the frame 802, the first sensor 804 having a first field of view. In this example, the sensor pod 800 also includes a second sensor 808 that has a second field of view, mounted at a second position 810 of the frame 802, the second position 810 being oriented relative to the first position 806 to overlap at least a portion of the first field of view with at least a portion of the second field of view. An example of this may be seen in Figure 2. In this example, the sensor pod 800 also includes a third sensor 812 having a third field of view, mounted at a third position 814 on the frame 802, where the third position 814 is oriented relative to the first position 806 and the second position 810 so that at least a portion of the second field of view overlaps with at least a portion of the third field of view. In this example, the sensor pod 800 also includes a fourth sensor 816 having a fourth field of view, detachably mounted at a fourth position 818 on the frame 802. In this example, the fourth position 818 is oriented relative to the first position 806 so that at least a portion of the first field of view overlaps with at least a portion of the fourth field of view. In this example, the sensor pod 800 also includes a fifth sensor 820 having a fifth field of view, mounted at a fifth position 822 on the frame. In this example, the fifth position 822 is oriented relative to frame 802 so as to overlap at least a portion of the fifth field of view with at least a portion of the fourth field of view.

[0059] In the example, the first sensor 804, the second sensor 808, and the third sensor 812 are of the first type, while the fourth sensor 816 and the fifth sensor 820 are of the second type. In the example, the first type of sensor is an imaging sensor, such as a camera. In the example, the second type of sensor is a LiDAR sensor. In the example, the fields of view for different types of sensors may be the same or different. For example, the field of view of the first sensor 804 may be different from the field of view of the third sensor 812. In the example, the field of view of the third sensor 812 is the same as the field of view of the first sensor 804 rotated 90 degrees.

[0060] In this example, the first sensor 804 is oriented with its optical axis in the first direction, the second sensor 808, which has an optical axis, is oriented in the second direction, and the third sensor 812, which has an optical axis, is oriented in the third direction, which is substantially opposite to the first direction. In this example, the second direction lies between the first and third directions, and the first and second fields of view are distinct from the third field of view.

[0061] Figure 8 also shows a portion of the cleaning system 824. In this example, the cleaning system 824 is for cleaning the sensing surface 826 of a sensor, for example, a first sensor 804 or a fourth sensor 816. In this example, the cleaning system 824 includes a nozzle 828 configured to apply a liquid to the sensing surface 826.

[0062] In the example, sensor pod 802 has a supply harness 830. In the example, the supply harness 830 provides power, control signals, and / or cleaning fluid from the vehicle to sensor pod 800. In the example, the supply harness 830 or another harness (not shown) provides data signals from the sensor to the vehicle. In the example, the supply harness includes a fluid and pressurized air connection for supplying fluid and pressurized air to the cleaning system and a power connection for supplying power to one or more of the sensors. In the example, sensor pod 800 is electrically connectable to the vehicle to transmit sensor data from the sensor to the vehicle's computing system and includes a sensor harness electrically connected to the sensor.

[0063] In the example, frame 802 may be made from different materials. For example, frame 802 may be made from metal (e.g., aluminum, steel, magnesium, or a combination thereof) or composite materials including carbon, Kevlar, resin, glass, or plastic. In the example, frame 802 is made from aluminum casting or magnesium-aluminum alloy. In the example, a cast frame provides manufacturing strength, rigidity, and repeatability at a lower cost than a frame fully machined from a billet. In the example, a mounting position on frame 802 has a mounting surface for supporting and oriented the sensor by aligning away from the mounting surface. In the example, the mounting surface is machined to give the reference surface sufficient tolerance to orient the sensor relative to each other and to the vehicle. In the example, a mounting position on frame 802 has a cast surface at the mounting position for supporting and oriented the sensor. In these examples, the cast surface does not require a secondary machining operation to give the reference surface. Rather, the cast surface is controlled during the casting process to give the reference surface sufficient tolerance to orient the sensor relative to each other and to the vehicle.

[0064] In this example, the sensor pod is asymmetrical. For example, the mounting interface for the sensor pod is located at a sixth position on the frame, where the sixth position is substantially opposite to the second position 810. In this example, the sensor pod mount protrudes from the side of the sensor pod 800 when coupled to the vehicle.

[0065] Exemplary pedestrian protection systems and techniques Safety is a critical factor in vehicles, such as autonomous vehicles. As part of safety, it is also important to provide protection to people in the environment in which the vehicle operates. For example, a vehicle may have a complex system to help prevent unintended contact with people in its environment. In situations where contact cannot be prevented, additional protection may be provided on the vehicle. For example, this type of pedestrian protection may be designed to limit injuries to pedestrians.

[0066] One current measure of pedestrian protection is the Head Injury Criteria ("HIC") score. This is a metric that determines the level of pedestrian protection provided by a vehicle. For example, a pedestrian protection system with an HIC score below 1000 is considered satisfactory in many jurisdictions. In this example, the HIC score is calculated by Equation 1.

[0067]

number

[0068] Here, a is the obtained head acceleration, t2-t1 ≤ 15 ms, and t2 and t1 are selected to maximize the HIC. The HIC is a measure of acceleration concentration as a proxy for force / energy applied over a time period between t1 and t2. In the example, the system may use one or more of the techniques described in Regulation (EC) No. 78 / 2009 of the European Parliament and the Council of 14 January 2009 (describes the head performance criteria (“HPC”)) and in the European New Vehicle Assessment Programme Pedestrian Test Protocol, Version 8.4, November 2017 (describes the HIC15 test), whose disclosures are incorporated herein by reference.

[0069] Figure 9 shows a perspective view of the sensor pod 800. In this figure, the sensor pod 800 is fitted with a trim / shell 900. In this example, the sensor pod 800 includes a collision structure 902 coupled to a frame 802. For example, the collision structure 902 may include an outer surface 904 configured to interface with a pedestrian during a collision. In this example, at least a portion of the outer surface 904 is positioned outside the sensor relative to the frame 802. In this example, the collision structure 902 also includes a collision energy absorption structure 906 positioned between the outer surface 904 and the frame 802. In this example, the collision energy absorption structure 906 is configured to absorb a portion of the energy transmitted from the impact through the outer surface 904.

[0070] In the example, the impact structure 902 and / or the impact energy absorption structure 906 may be made from a variety of materials, including, for example, nylon, glass-filled nylon, and / or glass-filled polypropylene. For example, the impact energy absorption structure 906 may be made from plastics (e.g., polyethylene terephthalate (PET or PETE or polyester), high-density polyethylene (HDPE), polyvinyl chloride (PVC), low-density polyethylene (LDPE), polypropylene (PP), polystyrene (PS), (ABS), etc.), polycarbonate, polyamide, and / or combinations thereof. In the example, the impact energy absorption structure 906 may be made from metals (e.g., aluminum, steel, magnesium, or combinations thereof) or composite materials including carbon, Kevlar, resin, glass, or plastic.

[0071] In the example, the sensor pod includes an inertial measuring unit ("IMU") or sensor. In the example, the vehicle receives data from the IMU indicating an impact on the sensor pod. In the example, the system may determine, at least partially, that the sensor pod is damaged, may be damaged, needs to be inspected, needs to be recalibrated, or needs to be replaced. In the example, the system may determine, at least partially, the impact force applied to a pedestrian or object during the collision.

[0072] In the example, data from the sensor pod's IMU is combined with data from the vehicle and / or other sensor pods. For example, data from the sensor pod's IMU showing high deceleration may not indicate a collision with the sensor pod when, for example, data from the vehicle and / or other sensor pods shows similar high deceleration. In the example, a large discrepancy between the data from the sensor pod and the data from the vehicle may indicate poor alignment and / or miscalibration of the sensor pod and / or IMU.

[0073] Figure 10 is a simplified diagram of the processes and techniques for managing energy from a collision through a pedestrian protection system. For example, Figure 10 shows several schematic diagrams of a collision structure 902 coupled to a frame 802. In the example, the collision structure 902 includes an outer surface 904 configured to interface with the pedestrian and a collision energy absorption structure 906 during a collision. Figure 10 shows an exemplary process 1000 for mitigating a collision with a pedestrian or other external object. For example, in operation 1002, the collision structure, e.g., collision structure 902, absorbs energy from the structure, e.g., the outer surface 904, and absorbs the collision through it. In the example, the outer surface 904 may be displaced in the direction of arrow 1004, e.g., into the sensor pod. In the example, the displacement of the outer surface 904 begins to absorb energy from the collision and direct it away from the pedestrian.

[0074] In operation 1006, the collision may transfer energy to the collision energy absorption structure 906, creating a stress concentration. In this example, the energy transferred to the collision energy absorption structure 906 may exceed an amount that causes the stress concentration to exceed a threshold, deforming a portion of the collision energy absorption structure 906 and absorbing energy from the collision. In this example, the collision energy absorption structure 906 includes a stress concentration zone 1008 during the collision. In this example, the stress concentration zone 1008 causes local plastic deformation of the collision energy absorption structure, thereby absorbing energy away from the collision and from the pedestrian. For example, this energy absorption reduces the acceleration concentration experienced by the pedestrian. In this example, the stress concentration zone 1008 is configured to produce local plastic deformation that is above a first collision force threshold and below a second collision force threshold. In this example, it is useful that the collision energy absorption structure is deformable enough to absorb enough energy from the collision to protect the pedestrian, but not so deformable that it causes unwanted damage during driving or minor collisions. Furthermore, it is useful that the collision energy absorption structure is not too rigid so as not to absorb enough energy from the collision in order to protect pedestrians.

[0075] In the example, the stress concentration zone 1008 is adapted using material properties and / or geometric features. For example, the cross-sectional thickness of the impact energy absorption structure 906 may be modified across the structure to locate and / or limit the stress concentration zone 1008 at a certain location, having a region to absorb a desired amount of energy. In the example, geometric features include added material, such as ridges, ribs, swells, and gussets; removed material, such as slots, holes, or recesses; and / or contours, such as fillets, curves, or radii. In the example, combinations are used to create complex shapes, such as I-beams, C-channels, tubes, and / or T-beams.

[0076] In the example, the second collision force threshold is associated with an HIC score of 1000. In the example, the second collision force threshold is associated with an HIC score of 650. In the example, the first collision force threshold is associated with an HIC score of 100. In the example, the first collision force threshold is associated with an HIC score of 300.

[0077] In this example, the outer surface 904 is configured to move substantially uniformly relative to itself during the collision. In this example, the collision energy absorption structure 906 absorbs most of the energy transferred during the collision.

[0078] In the example, the outer surface is configured to deform or bend relative to itself up to the maximum deformation. In the example, the maximum deformation is approximately 50% of the thickness of the outer surface's shape.

[0079] In the example, the outer surface is substantially convex or curved. In the example, this shape can impart structural stiffness to the weight-to-weight ratio. In the example, the outer surface is substantially convex in two orthogonal directions. In the example, the outer surface is curved in two directions, which can create a shell effect that imparts structural stiffness to the weight-to-weight ratio.

[0080] In the example, the collision energy absorption structure 906 connects the outer surface 904 to the frame 802. In the example, the collision energy absorption structure 906 provides support to the outer surface 904 in a direction substantially perpendicular to the direction of impact. For example, the collision energy absorption structure 906 may help support the outer surface in both vertical and lateral directions.

[0081] In the example, the impact energy absorption structure 906 has one or more of the following characteristics: thickness, cross-section, or planar shape, configured to give an impact load absorption profile. In the example, these characteristics are adapted to control stress concentration and deformation of the impact energy absorption structure 906. In the example, the impact energy absorption structure may be composed of a substantially homogeneous material or a composite material.

[0082] In the example, the impact structure 902 may provide protection through application to another sensor mounted on a frame, for example, frame 802, and coupled to frame 802 by a deformable fastener. For example, the sensor mounted on frame 802 may be a LIDAR sensor. In the example, the sensor is coupled to the frame through a deformable fastener having a stress concentration zone. In the example, the stress concentration zone of the deformable fastener causes local plastic deformation of the deformable fastener that absorbs energy from the impact. In the example, when the impact exceeds an impact force threshold, the deformable fastener releases the sensor. For example, when the impact is sufficiently large, the deformable fastener may allow it to change orientation and detach the sensor from the frame. In the example, when the sensor is released from the frame, if it is mounted near the top of the vehicle, it may pass without touching a pedestrian or obstacle. In the example, if the sensor does not pass without touching an obstacle, the deformable fastener absorbs enough energy to slow the relative velocity of the sensor enough to mitigate a secondary impact. In this example, the deformation of the deformable fastener absorbs energy, reducing the overall energy absorbed by the collision energy absorption structure 906 away from the sensor.

[0083] The collision structure 902 may also include a baffle structure. For example, Figures 8 and 9 also show a baffle structure 832 disposed between sensor 804 or sensor 808 and the outer shell 900. In this example, the baffle structure 832 includes a crumple zone configured to plastically deform during collision and absorb energy from the collision. In this example, this feature operates as well as and in conjunction with the outer shell 900, the outer surface 904, and the collision energy absorption structure 906.

[0084] Exemplary calibration systems and techniques The techniques described herein concern the calibration of sensors in a sensor pod and a vehicle without requiring infrastructure, for example, without reference markers. Generally, such calibration may refer to either an "external" calibration (determining one or more of the position or orientation of the sensor relative to some origin, e.g., another sensor, a system origin, etc.) or an "internal" calibration (determining one or more parameters about the sensor itself, e.g., focal length, center point, lens distortion model, etc.). An example of a system is an autonomous vehicle with a diverse sensor pod having diverse sensors (of various modalities), but any other system (e.g., a smartphone with an image sensor, a robotic manipulator with one or more sensor modalities, etc.) is contemplated. In one such example (i.e., the system is an autonomous vehicle), the autonomous vehicle may include sensors in the sensor pod configured to generate sensor data of the vehicle or the sensor pod's environment.

[0085] A vehicle or sensor pod may generate, or otherwise access, semantic segmentation information that identifies objects in the environment. For example, semantic segmentation information may classify sensor data, for instance, pixel by pixel or point by point, to associate each pixel or point with an object or object type. When a sensor is properly calibrated (for example, internally), sensor data and semantic segmentation information can be combined to generate an aligned representation of the environment. However, when a sensor is improperly calibrated, for example, with calibration errors, the sensor data and semantic segmentation information, when combined, can form an inaccurate or "blurred" representation of the environment. In real-world examples, when applied to autonomous vehicles, a sensor misalignment of an order of 0.5 degrees can result in the inability to reliably determine the lane an object 100 meters away is moving in, as this illustrates the need for extremely precise calibration. Therefore, while sensors are calibrated at installation to meet these stringent standards, it may be crucial to verify that the sensors remain properly calibrated. For example, sensors can be subject to environmental pollutants, degradation from normal operation and use, and other factors that can cause calibration errors in the sensors.

[0086] When installing sensor pods, it may be beneficial to take into account the calibration factors of the sensor pods. For example, in some cases, biases and / or calibration limitations may exist within a given sensor pod and may limit the compatibility of two sensor pods. For instance, a first sensor pod may have a first bias, and therefore, when paired on a vehicle with a second sensor pod having a second bias, the combination of sensor pods may, in some cases, not provide the desired sensor coverage. However, when the first sensor pod is paired with a third sensor pod having a third bias, the two sensor pods will provide the desired sensor coverage. Some cases may be extreme scenarios where various sensors or sensor pods on a vehicle fail, and the remaining sensors are used to safely control the vehicle. In such cases, for example, one or more of the sensor pods may not need to be removed and replaced before the vehicle enters use, so determining the compatibility of sensor pods before installation on a vehicle can reduce vehicle reprocessing and downtime.

[0087] In the examples described herein, calibration techniques may be used to verify and / or calibrate the calibration of one or more sensors mounted in a sensor pod and / or in a vehicle. In some implementations, sensors of one or more modalities may be mounted to capture sensor data, such as images, radar return, LiDAR return, time-of-flight data, etc., and may cover the field of view of the sensor pod. In the examples, various sensor pods may be mounted in an autonomous vehicle, providing an effective field of view covering 360 degrees around the autonomous vehicle, and each sensor may provide sensor data necessary to ensure the safe navigation of the vehicle with respect to objects in the vehicle's environment. In the examples, calibration techniques may include leveraging a reference or immutable object when calibrating sensors on a sensor pod. When using immutable objects, calibration techniques may include leveraging semantic segmentation information to determine calibration errors. More specifically, the technique may include determining an object in the semantic segmentation information and comparing it with features in the sensor data generated by the sensor relative to the object. For example, objects in semantic segmentation information may include immutable objects, which may not change or move over time and / or may include fixed and / or known objects with one or more attributes readily known, such as geometric attributes or reflective features. For example, immutable objects may include linear objects such as walls, columns, ceilings, corners, rooflines of structures or vehicles, buildings, and horizons. Immutable objects may also include flat surfaces such as signs, surfaces, walls, sides of buildings, machinery, and sides of trailers or heavy vehicles. In another example, road signs may be known to be retroreflective. In yet another example, immutable objects may have other attributes that are known or can be known. In at least other examples, such immutability may correspond to the features and / or attributes of the object, rather than to the object itself.As a non-limiting example of such things, a tractor trailer can move but has a flat surface and a straight line at the intersection of such a surface. Similarly, a traffic signal pole changes its state (red, yellow, green, etc.) but may have the same relative position, orientation, geometric characteristics, etc.

[0088] In some implementations, features can be extracted from sensor data for comparison with objects in semantic segmentation information. As a non-restrictive example, edges can be detected in sensor data, for example, using edge detection algorithms in images or using depth discontinuities in 3D sensor data. Such edges can be investigated to determine whether they are linear, as expected from a priori knowledge of the invariant object. In other implementations, for example, points with depth measurements can be identified as being associated with a planar surface in the invariant object. Those points can then be investigated to check for coplanarity. In some examples, features can be determined from destrained data, for example, using sensor parameters, such as intrinsicity. In at least some examples, errors can be determined with respect to distorted space (e.g., how far a point is from an expected distorted line or plane) or destrained space (e.g., measures of collinearity or coplanarity).

[0089] As described above, the sensed features must be aligned in close proximity to objects from semantic segmentation information in a well-calibrated sensor. On the other hand, misalignment may indicate calibration errors in the sensor. Therefore, according to the implementations described herein, the technique can quantify how much features in the sensor data deviate from their predicted values, for example, how much they deviate from the collinearity or coplanarity expected from invariant object attributes. In some examples, misalignment can also be quantified, for example, as the distance between a point or pixel in the sensor data and the prediction made by the invariant object determined from the semantic segmentation information.

[0090] In some cases, the calibration error may be compared to a threshold error, for example, to determine whether any corrective action must be taken to determine if the error is acceptable. Corrective actions may include taking the sensor offline, taking the sensor pod offline, taking the vehicle offline, recalibrating the sensor, or controlling the vehicle when sensor data is removed from the sensor. In some cases, the calibration data may be used to determine updated calibration information, and subsequently acquired sensor data may be calibrated using the updated calibration information.

[0091] The calibration techniques described herein can improve the functionality of computing devices by providing a framework for determining optimal calibration for sensors on sensor pods of autonomous vehicles, such as arrays of cameras. By calibrating one or more cameras using the calibration techniques described herein, the cameras can generate data representing the environment with high accuracy and precision while providing the desired field of view coverage for the vehicle. For example, a sensor pod calibrated in this manner can provide more consistent sensor coverage during operation, which ultimately leads to better safety outcomes during driving and reduces the amount of vehicle reprocessing that needs to be achieved. These and other improvements to the functionality of computing devices are described herein.

[0092] The calibration techniques described herein also represent an improvement over conventional calibration and / or calibration verification. For example, in the past, calibration techniques often required the sensors to be installed and activated on the vehicle. For example, some techniques require the sensors to be installed, activated, and tested on the vehicle. Such conventional techniques are unavoidable, requiring intensive time and integration efforts with the vehicle offline for implementation. In contrast, the techniques described herein may enable the calibration of sensors at the sensor pod level, for example, which can be performed separately from the vehicle while the vehicle is still in use. As an addition or alternative, the techniques described herein may enable the calibration of the vehicle with a new sensor pod installed for faster and more efficient calibration. For example, the technique may enable the vehicle to calibrate the sensors of a new sensor pod using a reduced set of variables. Furthermore, the technique may enable the vehicle to calibrate the sensors of a new sensor pod, for example, because the vehicle knows that a calibration solution exists, since the sensor pods have been installed in locations determined to be compatible with each other. Therefore, the techniques described herein represent a significant improvement over conventional calibration methods.

[0093] The methods, apparatus, and systems described herein can be implemented in several ways. Exemplary implementations are given below with reference to the following figures. Although described in the context of autonomous vehicles, the methods, apparatus, and systems described herein can be applied to a variety of systems that require sensor calibration before and / or during the use and / or verification of such calibration. The techniques described herein are not limited to autonomous vehicles. In another example, the methods, apparatus, and systems may be used in an aviation or navigation context. Furthermore, the techniques described herein can be used with real data (e.g., captured using one or more sensors), simulated data (e.g., generated by a simulator), or any combination thereof.

[0094] Figures 11-13 are flowcharts illustrating exemplary methods and processes involved in the calibration and installation of sensors and sensor pods on autonomous vehicles. For convenience and ease of understanding, the methods shown in Figures 11-13 are described in relation to one or more of the vehicles, sensor pods, and / or systems shown in Figures 1-10. However, the methods and processes shown in Figures 11-13 are not limited to being carried out using the vehicles, sensor pods, sensors, and / or systems shown in Figures 1-11, but may be implemented using any of the other vehicles, sensor pods, and / or systems described herein, as well as vehicles, sensor pods, and / or systems not described herein. Furthermore, the vehicles, sensor pods, and / or systems described herein are not limited to carrying out the methods and processes shown in Figures 11-13.

[0095] In some examples, the system may use one or more of the techniques described in U.S. Patent Application No. 15 / 820,245 filed November 21, 2017 and / or U.S. Patent Application No. 16 / 297,127 filed March 8, 2019, which are incorporated herein by reference, to calibrate individual sensors or sensors on the vehicle.

[0096] Figure 11 shows an exemplary process 1100 for calibrating a vehicle sensor pod. In operation 1102, the calibration system may receive sensor data of the sensor pod's environment. In the example, the sensor pod may be mounted on a stand and computing system to collect and / or process sensor data. In the example, the stand may be fixed or freely movable. In the example, the sensor pod may be held by hand during data acquisition. In the example, the sensors of the sensor pod may be cameras or other sensors configured to capture data, e.g., images. In the example, the images may be distorted by the intrinsic features of the sensor, e.g., fisheye or wide-angle lenses. The sensors may be other types of cameras, e.g., linear cameras, etc., or the sensors as a whole may be sensors of different modalities. In a non-limiting example, the sensors may be LiDAR sensors or time-of-flight sensors that generate depth information related to the return of light, etc.

[0097] In the example, the environment includes one or more invariants within a 90-degree field of view. In the example, the test mount may have indexed orientations, where data may be collected from various sensors in diverse indexed orientations of one or more invariants. In this example, fewer invariants may be required to calibrate the sensors in the sensor pod.

[0098] In the example, the sensor data includes image data of multiple objects in the environment, such as a reference point, a wall, a corner of a building, a vehicle, a machine, a pole, and other features. As you can see, the image data may include sensed data of invariant, variable, static, dynamic, and / or arbitrary features within the sensor's field of view.

[0099] In operation 1104, the process may include receiving and / or generating semantic segmentation information about the environment (for example, based at least in part on sensor data acquired in 1102). In the example, the semantic segmentation representation may show classification information about a portion of the environment. More specifically, the representation may include graphical object representations of different objects or features in the environment based on the classification of those objects. For example, object representations may include, in particular, walls, corners, ceilings, poles, and a machine. In the example, each of the respective object representations may visually identify pixels that have the same marking or classification. Thus, the wall representation includes pixels that have the marking or classification "wall," the ceiling representation includes pixels that have the marking or classification "ceiling," and so on. Thus, for example, the semantic segmentation information may include per-pixel classification information. As described above, Figure 11 is illustrated with respect to a camera where the sensor generates an image, but other sensor modalities (e.g., a 3D sensor) are also contemplated. Similarly, semantic segmentation information can be generated from data from one or more sensor modalities, which may or may not include image sensors.

[0100] In some examples, semantic segmentation information may include information about invariant objects. As used herein, “invariant object” may be an object having known characteristics or attributes, such as linear ends, planar surfaces, or known orientations. In some examples, an invariant object may be an object that is fixed in the environment, for example, an object that is generally expected to remain the same or not move. For example, an invariant object may be a topographic feature, such as a road surface, a sidewalk surface, or some trees; a fixed object, such as a building, a road sign, a streetlamp, a fire hydrant, a guardrail; or another generally fixed object or feature having a known structure and / or orientation. In some examples of this disclosure, the horizon may be an invariant object, for example. In further examples, an invariant object may include a non-stationary object having some known characteristics or attributes. For example, the trailer of a tractor trailer may be an invariant object because it may have known linear ends and / or planar sides. Implementations of this disclosure may use semantic segmentation information about invariant objects. In other implementations, semantic segmentation information may include additional semantic segmentation information for non-immutable objects, such as, but not limited to, information about vehicles, pedestrians, and cyclists. However, the techniques described herein may only consider data related to immutable objects or criteria.

[0101] In operation 1106, process 1100 includes identifying sensor data corresponding to one or more invariant objects in semantic segmentation information. For example, image data corresponding to one of the invariant objects may represent features known to be associated with the invariant object. For example, image data associated with a wall may have, for example, a vertical end at the corner of the wall and / or a horizontal end at the bottom of the wall, ceiling, etc. Similarly, image data associated with a pole may have, for example, two vertical ends in the lateral range of the pole, since the pole is known to be relatively straight and / or have a vertical orientation. However, as described above, many factors can affect the actual calibration of the sensor. For example, manufacturing and / or assembly tolerances related to the sensor, sensor pod, vehicle, sensor pod mount on the vehicle, etc., may result in an inappropriate calibration of the sensor for reasons unknown. Furthermore, the sensor may become uncalibrated through normal operation due to environmental fluctuations such as vibrations, temperature, and / or other factors related to the vehicle, for example. The techniques described herein are useful in identifying misalignments caused by these and other sources. As an addition or alternative, semantic segmentation comparisons of different sensor modalities (whether they contain invariant objects) may be made. In such examples, known calibrations may be used to transform data in one modality into a reference frame relevant to the other. Differences between the relevant modalities (and / or associated gross statistics) may be used to infer calibration errors (whether intrinsic or exogenous). In at least some examples, semantic information related to sparse data (e.g., LiDAR) may be projected onto a denser sensor modality (e.g., imaging) using the previous calibration information for comparison.

[0102] In the example, comparing sensor data, e.g., image data, with predictions of image data corresponding to invariant features collected from semantic segmentation information can reveal calibration errors. In the example, the system can identify points associated with invariant objects based on semantic segmentation information. For example, a pole can be an invariant object because it includes (vertical) linear ends or shapes.

[0103] In operation 1108, the process may include determining a calibration error for the sensor. For example, the techniques described herein may determine the error based on the distance between sensor data and expected values ​​associated with those points (based on known attributes of an immutable object). In at least one example, such an error may comprise the distance between individual points based on semantic segmentation information and the expected shape of an immutable object. The distance may represent the calibration error. In some examples, the distance may be a Euclidean distance measured in pixels between each point and the expected shape. In some examples, the distance may depend on other characteristics of the sensor (for example, a fisheye camera may have high radial distortion, and the Euclidean distance determined for the error may be inversely scaled with respect to the radius). In other examples, an arc may be fitted to the points, and the arc is compared to the expected shape to determine the calibration error. In some implementations, if the error determined in operation 1108 is within some threshold error, for example, less than or equal to it, the sensor may be determined to be properly calibrated. However, if one or more errors are outside a certain threshold error, for example, above it, the sensor may be determined to have an inaccurate calibration.

[0104] In some implementations, semantic segmentation information may include markings for all pixels in a region, and thus a pixel may be determined as the outermost pixel having a "pole" marking or classification. In some implementations, lines may be fitted to pixels, and these lines represent markings in the semantic segmentation information.

[0105] In the example, the process involves determining calibration errors for the various sensors in the sensor pod. In the example, the various sensors are calibrated individually or in parallel. In the example, the immutable object may be the same object or different objects. In the example, the immutable object may be located where the fields of view of two or more sensors overlap.

[0106] In operation 1110, the process may include determining a calibration factor for the sensor pod. As described with respect to operation 1108, various sensor calibration errors may be determined. In the example, individual sensor errors may be combined to determine a calibration factor for the sensor pod. In the example, the calibration factor may contain a variety of values. For example, the calibration factor may contain a value for each sensor representing the calibration error of each sensor. In the example, the calibration factor for the sensor pod may be used in various processes. For example, the calibration factor may be used to calibrate the sensor pod when it is installed in a vehicle. As an addition or alternative, the calibration factor may be used to match sensor pods with other compatible sensor pods prior to installation on a vehicle. In some examples, such a calibration factor may be used in further processing so that a calibration-dependent algorithm can reduce one or more components of the associated miscalibrated sensors. In various examples, determining a miscalibrated sensor may involve comparing a single sensor to a variety of other sensors (e.g., two or more additional sensors in turn) to determine the reliability of the miscalibration.

[0107] Figure 12 shows an exemplary process 1200 for calibrating a sensor pod on a vehicle when the sensor pod is installed on the vehicle. In operation 1202, the system may receive sensor data of the environment of the sensor on the sensor pod. In the example, the sensor may receive sensor data in the same way as the process described with respect to Figure 11. In the example, the sensor data may include image data of multiple objects in the environment, such as a reference point, a wall, a corner of a building, a vehicle, a machine, a pole, and other features. As understood, the image data may include sensed data of invariant, variable, static, dynamic, and / or arbitrary features in the sensor's field of view.

[0108] In operation 1204, the process may include receiving and / or generating semantic segmentation information about the environment (for example, based at least in part on sensor data acquired in 1202). In the example, the semantic segmentation representation may show classification information about a portion of the environment. In the example, the sensor may receive or generate semantic segmentation information in the same way as the process described with respect to Figure 11.

[0109] In operation 1206, process 1200 includes identifying sensor data corresponding to one or more immutable objects in the semantic segmentation information. In an example, image data corresponding to one of the immutable objects may represent features known to be associated with the immutable object. For example, image data associated with a wall may have, for example, a vertical end at a corner of the wall and / or a horizontal end at the bottom of a wall, ceiling, etc. In an example, the sensor may identify immutable objects in a similar manner to the process described with respect to Figure 11.

[0110] In operation 1208, the process may include determining a calibration error for the sensor. For example, the technique described herein may determine the error based on the distance between sensor data and expected values ​​associated with those points (based on known attributes of immutable objects). In the example, the sensor may identify immutable objects in a similar manner to the process described with respect to Figure 11.

[0111] In operation 1210, the process may include determining the calibration error for another sensor of the sensor pod. In the example, the calibration factor for the sensor pod is based on the per-sensor calibration error of the sensor pod relative to the interface of the sensor pod to the sensor pod, e.g., a test stand or a vehicle mount. In the example, the calibration error determined in operation 1208 is based on the calibration error of the sensor pod's sensor relative to the vehicle. In this implementation, the sensor calibration error is combined with the sensor pod calibration factor to determine the estimated calibration error for the other sensors of the sensor pod relative to the vehicle. In the example, the estimated calibration error may be verified by the vehicle. In the example, the estimated calibration error gives the vehicle's system a precise starting point for verifying or refining the calibration of the other sensors. In the example, this ensures a closed solution when resolving the calibration error in some technique. As described above, the calibration factor may also be used to match sensor pods with other compatible sensors prior to installation on the vehicle.

[0112] Figure 13 shows an exemplary process 1300 for determining, prior to installation, whether a sensor pod is compatible with, for example, another sensor pod on a vehicle. In the example, sensor pod compatibility with another sensor pod is whether the effective fields of view of each sensor pod are compatible with each other. In the example, the effective fields of compatible sensor pods give a field of view overlap that exceeds a field of view overlap threshold. In operation 1302, the system may receive sensor data of the sensor environment on the sensor pod and identify invariant objects. In the example, the sensor may receive sensor data, generate semantic segmentation data, and identify invariant objects, similar to the processes described with respect to Figures 11 and 12. In the example, the sensor data may include image data of multiple objects in the environment, such as a reference point, a wall, a corner of a building, a vehicle, a machine, a pole, and other features. As understood, the image data may include sensed data of invariant, variable, static, dynamic, and / or arbitrary features in the sensor's field of view. In the example, the process may identify image data corresponding to one of the immutable objects that may represent features known to be associated with immutable objects. For example, image data associated with a wall may have, for example, a vertical end at the corner of the wall and / or a horizontal end at the bottom of a wall, ceiling, etc.

[0113] In operation 1304, the process may include determining a calibration factor for the sensor pod. In the example, calibration errors for various sensors may be determined. In the example, individual sensor errors may be combined to determine a calibration factor for the sensor pod. In the example, the calibration factor may include a variety of values. For example, the calibration factor may include a value for each sensor representing the calibration error of each sensor. In the example, the calibration factor for the sensor pod may be used in various processes. For example, the calibration factor may be used to calibrate the sensor pod when it is installed in a vehicle. As an addition or alternative, the calibration factor may be used to match sensor pods with other compatible sensor pods before installation on a vehicle, for example, by determining whether the field of view of each sensor pod gives the desired field of view overlap. In the example, the sensor may determine its calibration factor in the same way as the process described with respect to Figure 11.

[0114] In operation 1306, the system may receive the calibration factor of another sensor pod. For example, a second sensor pod may be considered to be for pairing with a first sensor pod, where the calibration factor of the first sensor pod is determined in operation 1304, and the calibration factor for the second pod is received in this operation. In the example, the second sensor pod may be a sensor pod already installed on the vehicle. In the example, the second sensor pod may be a sensor pod that is not yet installed on the vehicle.

[0115] In operation 1308, the process determines whether a first sensor pod is compatible with a second sensor pod by determining, for example, whether the field of view of the first sensor pod, when combined with the field of view of the second sensor pod, provides the desired field of view overlap. In the example, the decision is based at least in part on the calibration factors of the first sensor pod and the calibration factors of the second sensor pod. In the example, the first and second sensor pods may be compatible with each other in normal use regardless of their calibration factors. However, in some examples, the sensor pods may not provide the desired sensor redundancy or field of view overlap in certain situations. For example, if a sensor or sensor pod fails or goes offline, the remaining sensors or sensor pods may require a threshold level of sensor coverage. In this example, the process uses the calibration factors of each sensor pod to determine whether the first and second sensor pods would provide such coverage if some sensors are lost. While these examples describe two sensor pods for clarity, the process may consider three, four, or more sensor pods when determining compatibility. For example, the process may determine whether a first sensor pod fits three other sensor pods. For example, the process may determine whether the effective field of view of the first sensor pod and the three other sensor pods provides the desired field of view overlap. In a non-limiting example, the process may determine whether a replacement sensor pod fits three sensor pods already installed on the vehicle. For example, the process may determine whether the effective field of view of the already configured first sensor pod and the three other sensor pods provides the desired field of view overlap.

[0116] If, in operation 1308, the process determines that the sensor pod is unsuitable, the process returns to operation 1306 to receive the calibration factor for another sensor pod to be evaluated. For example, the process determines that the field of view of the sensor pod does not provide the desired field of view overlap.

[0117] In operation 1308, if the process determines that a sensor pod is suitable, for example, has a suitable field of view, then in operation 1310, the two sensor pods are paired. In the example, when the sensor pods are paired, an indication that the sensor pods are suitable, for example, have a suitable field of view, may be included in the database or given to the operator. In the example, if the two sensor pods are not installed on the vehicle, the sensor pods may be stored and later installed together on the vehicle. In the example, if one of the sensor pods is already installed on the vehicle, the other sensor pod may be reserved until a replacement sensor pod is needed.

[0118] In operation 1312, the sensor pod is installed on the vehicle together with other sensor pods. In the example, the sensor pod may be a replacement sensor pod. In the example, the sensor pod may be grouped with additional compatible sensor pods and installed on the vehicle simultaneously.

[0119] Figure 14 shows a block diagram of an exemplary system 1400 for implementing the techniques described herein. In at least one example, system 1400 may include a sensor pod platform 1402 and a control system or a vehicle, which may be the same vehicle 100 described above with reference to Figures 1 to 4.

[0120] The sensor pod platform 1402 may include a constrained test mount 1404, an unconstrained test mount 1406, or a vehicle 1408, and may include a computing device 1410, one or more sensor systems 1412, for example, one or more sensor pods, and one or more communication connections 1414.

[0121] The computing device 1410 may include one or more processors 1416 and a memory 1418 communicatively coupled to one or more processors 1416. In the illustrated example, the sensor pod platform 1402 may be a test mount or an autonomous vehicle, but the autonomous vehicle may be any other type of vehicle which may or may not be autonomous. Furthermore, as described herein, the techniques described herein may be any device which has sensors and has access to (or is not able to generate) semantic segmentation information about the sensor's environment, as described herein. In the illustrated example, the memory 1418 of the computing device 1410 stores a localization component 1420, a perception component 1422, a planning component 1424, one or more system controllers 1426, and a calibration component 1428. As shown in the figure, the perceptual component 1422 may include the semantic segmentation component 1430, and the calibration component 1428 may include the feature detection component 1432, the error determination component 1434, and the strain removal component 1436. Memory 1418 may also store one or more maps 1438. Although shown in Figure 14 as residing in memory 1418 for illustrative purposes, it is intended that several features, including the calibration component 1428, the semantic segmentation component 1430, the maps 1438, and / or other components, may be accessible to the sensorpod platform 1402 (for example, stored remotely) as an addition or alternative.

[0122] In at least one example, the localization component 1420 may include the ability to receive data from the sensor system 1412 to determine the location of the sensor pod platform 1402. For example, the localization component 1420 may include, request, and / or receive a 3D map of the environment from, for example, map 1438, and sequentially determine the location of the sensor pod platform 1402 within the map. In some examples, the localization component 1420 may utilize SLAM (Simultaneous Localization and Mapping) or CLAMS (Simultaneous Calibration, Localization, and Mapping) to receive image data, LIDAR data, radar data, SONAR data, IMU data, GPS data, wheel encoder data, etc., to accurately determine the location of the sensor pod platform 1402. As described herein, the localization component 1420 may receive calibrated sensor data, for example, sensor data within a threshold acceptance range.

[0123] In some examples, the perceptual component 1422 may include functions for performing object detection, segmentation (e.g., semantic segmentation via functions provided by the semantic segmentation component 1430), and / or classification. In some examples, the perceptual component 1422 may provide processed sensor data indicating the presence of entities adjacent to the sensor pod platform 1402 and / or the classification of entities as entity types (e.g., cars, pedestrians, cyclists, animals, trees, road surfaces, curbs, sidewalks, streetlights, signs, unknowns, etc.). In additional and / or alternative examples, the perceptual component 1422 may provide processed sensor data indicating one or more characteristics related to the detected entities and / or the environment in which the entities are located. In some examples, characteristics related to the entities may include, but are not limited to, x-position (global position), y-position (global position), z-position (global position), orientation, entity type (e.g., classification), entity velocity, entity range (e.g., size), etc. Environmental characteristics may include, but are not limited to, the presence of other entities in the environment, the state of other entities in the environment, the time of day, day of the week, season, weather conditions, and the indication of darkness / light.

[0124] As described above, the perceptual component 1422 may also include a semantic segmentation component 1430. In some implementations, the semantic segmentation component 1430 may cause a computing device 1410 to perform semantic segmentation on image data, LiDAR data, and / or other sensor data generated by the sensor system 1412 to determine classifications or markings related to pixels, points, or other parts of sensor data. In some embodiments, the semantic segmentation component 1430 may include one or more machine learning algorithms directed to identify image data and segment it into semantic categories. For example, the semantic segmentation component 1430 may include a convolutional neural network (CNN) configured to perform semantic segmentation on an image or other sensor data and / or determine a pixel classification probability distribution for pixels in an image, but any other form of semantic segmentation on an image is intended. In at least some examples, the semantic segmentation component 1430 may determine classification or marking for data components including, but not limited to, cars, trucks, bicycles, motorcycles, pedestrians, particulate matter, buildings, road signs, streetlights, signs, trees, shrubs, and further including navigable surfaces, free space, drivable surfaces, etc. Furthermore, various classifications may be identified as immutable objects having several known attributes, features, or orientations. As described herein, semantic segmentation information may separately classify one or more immutable objects that may include lines, planes, or other known features.

[0125] In general, the calibration component 1428 may include functions for verifying calibration and / or calibrating one or more sensors operating with respect to one or more of the sensor pod platform 1402, for example, the sensor system 1412. For example, the calibration component 1428 may detect improperly calibrated sensors, schedule calibration routines, and send calibration data to one or more other components of the sensor pod platform 1402 that utilize data generated by one or more sensors.

[0126] The feature detection component 1432 can analyze image data to determine features, such as edges, planes, etc. Such image features may correspond to, for example, AKAZE, BRISK, SURF, SIFT, ORB, BRIEF, FAST, FREAK, embeddings, etc. In another example, the feature detection component 1432 can apply the Canny edge detection algorithm to detect edges in images captured by a camera mounted on the sensor pod platform 1402. In yet another example, the feature detection component 1432 can identify edges based on, for example, LiDAR with depth information, time of flight, and / or discontinuities in depth during return. For example, in the implementation described herein, edges detected by the feature detection component 1432 can be compared with semantic segmentation information to determine, for example, whether the detected edges align with the divisions between object classifications identified from semantic segmentation information.

[0127] The error determination component 1434 may include, for example, a function for determining misalignment of features relative to expected features associated with an object identified in semantic segmentation information generated by the semantic segmentation component 1430, such as features detected by the feature detection component 1432. For example, the error determination component may identify the distance between a point in the sensor data corresponding to a feature and a line, plane, or other feature representing an expected feature or attribute of an object that the semantic segmentation information indicates is associated with that point. In instances of edges or planes, the error determination component 1434 may include a function for fitting the line or plane to the sensed data and determining the error of the fitted line / plane relative to the expected return based on the semantic segmentation information, such as an offset.

[0128] The destraining component 1436 may include a function to destrain the sensor's return for comparison with the attributes of immutable objects and / or immutable features, for example, for destraining distorted image data. In some examples, the destraining component 1436 may store calibration data related to one or more sensors of the sensor pod platform 1402. In some examples, the calibration data may include information about the intrinsic nature of the sensors. In some examples, the calibration data may include, but is not limited to, any external and / or intrinsic information related to one or more sensors, including calibration angle, mounting position, height, orientation, yaw, tilt, pan, timing information, lens strain parameters, transmission medium parameters, etc. Furthermore, the calibration component 1428 may store a log of some or all of the calibration operations performed, such as the time elapsed since the most recent calibration. As mentioned above when describing the example in Figure 11, calibration error / verification may be performed using the techniques described herein while explicitly destraining the sensor data. In the example, invariant features, such as lines associated with poles, may be distorted in the image's distortion space. In at least some examples, optimization may be performed with respect to assumed calibrations in order to minimize errors and discover improved calibrations.

[0129] In some examples, for instance, the calibration component 1428, which performs error determination component 1434, can determine whether the calibration error is significant enough to affect the continued use of the sensor. In a non-limiting example, if the error is equal to or exceeds a threshold error for the sensor, the calibration component 1428 can communicate the sensor failure to other components of the sensor pod platform 1402. For example, the positioning component 1420, the sensing component 1422, and / or the planning component 1424 may perform functions when an improperly calibrated sensor is excluded. Other functions may be similarly limited. In other examples, the calibration component 1428 can, for example, when the platform is a vehicle, cause the system controller 1426 to stop the sensor pod platform 1402 until sensor calibration can be performed, and / or send information to a remote source, such as a remote operator, to notify about sensor misalignment and / or request a command to continue.

[0130] In some examples, the calibration component 1428 may also include a function for determining a correction function to correct calibration errors related to the sensor system 1412. For example, the calibration component 1428 can use the errors determined by the error determination component 1434 to adjust the sensor data. For example, to correct the sensor data using a correction function, the calibration component 1428 may consider data from multiple frames, returns, etc., and may consider miscalibrations for several different invariant objects. For example, identifying that points are not collinear may be sufficient to identify a calibration error, but correcting the error may not be achieved based on those points alone. In implementations, errors related to several different invariant objects and / or invariant features may be required to determine the calibration function. Furthermore, the relationships between invariant objects may also be considered when determining the calibration function. As a non-limiting example, when two invariant objects have planes that are expected to be perpendicular, the calibration component 1428 may better develop the calibration function by, for example, minimizing errors across a variety of degrees of freedom. However, in further embodiments, additional information may be required to accurately determine the calibration function.

[0131] In some cases, some or all aspects of the components described herein may include any model, algorithm, and / or machine learning algorithm. For example, in some cases, the components in memory 1418 (and memory 1446 described below) may be implemented as a neural network.

[0132] As described herein, an exemplary neural network is a biologically suggested algorithm that passes input data through a series of connected layers to produce an output. Each layer in the neural network may also comprise another neural network, or any number of layers (whether convolutional or not). As can be understood in the context of this disclosure, a neural network may utilize machine learning, which may refer to a broad class of such algorithms in which the output is generated based on learned parameters.

[0133] Although described in the context of neural networks, any type of machine learning may be used in accordance with this disclosure. For example, machine learning algorithms are not limited to, but include regression algorithms (e.g., ordinary least squares regression (OLSR), linear regression, logistic regression, stepwise regression, multivariate adaptive regression splines (MARS), locally estimated scatterplot smoothing (LOESS)), instance-based algorithms (e.g., ridge regression, least absolute shrinkage and selection operator (LASSO), elastic networks, least-angle regression (LARS)), decision tree algorithms (e.g., classification and regression tree (CART), iterative dichotomiser 3 (ID3), chi-squared automatic interaction detection (CHAID), decision stock, conditional decision tree), and Bayesian algorithms (e.g., naive Bayes, Gaussian naive Bayes, multinomial naive Bayes, average one-dependence estimator). estimators (AODE), Bayesian belief networks (BNN), Bayesian networks), clustering algorithms (e.g., k-means, k-means algorithm, expectation maximization (EM), hierarchical clustering), correlation rule learning algorithms (e.g., perceptron, backpropagation, Hopfield network, radial basis function network (RBFN)), deep learning algorithms (e.g., deep Boltzmann machine (DBM), deep belief networks (DeepBelief Networks (DBN), Convolutional Neural Networks (CNN), Stacked Autoencoders), Dimensionality Reduction Algorithms (e.g., Principal Component Analysis (PCA), Principal Component Regression (PCR), Partial Least Squares Regression (PLSR), Summon Mapping, Multidimensional Scaling (MDS), Projection Tracking, Linear Discriminant Analysis (LDA), Mixture Discriminant Analysis (MDA), Quadratic Discriminant Analysis (QDA), Flexible Discriminant Analysis (FDA)), Ensemble Algorithms (e.g., Boosting, Bootstrap Aggregation (Bagging), AdaBoost, Stack Generalization (Blending), Gradient Boosting Machines (GBM), Gradient Boosted Regression Trees) This may include Trees (GBRT), Random Forest, SVM (Support Vector Machine), supervised learning, unsupervised learning, and semi-supervised learning.

[0134] Examples of additional architectures include neural networks such as ResNet30, ResNet101, VGG, DenseNet, and PointNet.

[0135] Map 1438 may be used by sensorpod platform 1402 to determine its position and / or navigate within the environment, for example, when the platform is a vehicle. For the purposes of this description, the map can be any number of data structures modeled in two, three, or n dimensions, capable of providing information about the environment, such as topology (such as intersections), streets, mountain ranges, roads, terrain, and the environment itself. In some cases, the map may include, but is not limited to, texture information (e.g., color information (e.g., RGB color information, Lab color information, HSV / HSL color information), intensity information (e.g., LiDAR information, radar information, etc.), spatial information (e.g., image data projected onto a mesh, individual "surfaces" (e.g., polygons associated with individual colors and / or intensity)), and reflectivity information (e.g., reflection information, retroreflectivity information, BRDF information, BSSRDF information, etc.). In one example, the map may include a three-dimensional mesh of the environment. In some cases, the map may be stored in a tile format, so that individual tiles of the map represent distinct parts of the environment and can be loaded into working memory as needed. In some examples, map 1438 may include at least one map (e.g., an image and / or a mesh). The sensor pod platform 1402 may be controlled at least in part based on map 1438. That is, map 1438 may be used with the localization component 1420, the perception component 1422, the planning component 1424, and / or the calibration component 1428 to determine the position of the sensor pod platform 1402 for navigating within the environment, identify objects in the environment, and / or generate routes and / or trajectories. Furthermore, as described herein, map 1438 may include semantic segmentation information about immutable objects in the environment.As a non-limiting example, semantic segmentation information included in a map may be used in addition to, or instead of, semantic segmentation information generated by the semantic segmentation component 1430. In some examples, one or more of the maps 1438 may include a set of semantic segmentation information generated using sensor data generated by the sensor pod platform 1402 and / or one or more additional vehicles.

[0136] In at least one example, the sensor system 1412 could include a LiDAR sensor, a radar sensor, a time-of-flight sensor, an ultrasonic transducer, a SONAR sensor, a position sensor (e.g., GPS, compass, etc.), an inertial sensor (e.g., an inertial measurement unit, accelerometer, magnetometer, gyroscope, etc.), a camera (e.g., RGB, IR, intensity, depth, time-of-flight, etc.), a microphone, a wheel encoder, an environmental sensor (e.g., a temperature sensor, humidity sensor, light sensor, pressure sensor, etc.), and so on. The sensor system 1412 could include various instances of each of these or other types of sensors. For example, the LiDAR sensor (and / or radar sensor) could include individual LiDAR sensors (or radar sensors) located on the edges, front, back, sides, and / or top of the sensor pod platform 1402. In another example, the camera sensor could include various cameras positioned at various locations on the outside and / or inside of the sensor pod platform 1402. The sensor system 1412 can provide input to the computing device 1410. As an addition or alternative, the sensor system 1412 can send sensor data to one or more remote computing devices via one or more networks 1440 at a specific frequency, after a predetermined time period has elapsed, or in near real-time.

[0137] In some examples, the sensor system 1412 may be an active sensor system that includes controls for actively adjusting its parameters, for example. For example, some cameras may have an adjustable shutter speed or exposure time. Similarly, time-of-flight sensors, LiDAR sensors, radar sensors, etc., may have actively adjustable intensity and / or gain attributes. In some implementations, semantic segmentation information may be further used to adjust one or more settings of the sensor. For example, when semantic segmentation information identifies an object of a certain class type in the sensor's environment, the sensor may be adjusted to optimize the sensing of that object. For example, when some objects that are expected to have a certain color or brightness are identified from the semantic segmentation information, the intensity of the emitted light may be adjusted to optimize the sensing.

[0138] The processor 1416 of the sensorpod platform 1402 may be any suitable processor capable of processing data and executing instructions to perform operations as described herein. For example, but not limited to, the processor 1416 may comprise one or more central processing units (CPUs), graphics processing units (GPUs), or any other device or part of a device that processes the electronic data for conversion into other electronic data that can be stored in registers and / or memory. In some examples, integrated circuits (e.g., ASICs), gate arrays (e.g., FPGAs), and other hardware devices may also be considered processors, insofar as they are configured to implement encoded instructions.

[0139] Memory 1418 is an example of a non-temporary computer-readable medium. Memory 1418 can store an operating system and one or more software applications, instructions, programs, and / or data to implement the methods and functions resulting from various systems described herein. In various implementations, memory may be implemented using any suitable memory technology, such as static random-access memory (SRAM), synchronous dynamic RAM (SDRAM), non-volatile / flash-type memory, or any other type of memory capable of storing information. The architectures, systems, and individual elements described herein may include many other logical, programmatic, and physical components, and those shown in the accompanying figures are merely examples relevant to the description herein.

[0140] While Figure 14 shows a distribution system, in alternative examples, the components of the sensorpod platform 1402 may be associated with remote computer devices accessible via the network 1440. For example, the sensorpod platform 1402 can send sensor data to one or more computing devices 1442 via the network 1440. In some examples, the sensorpod platform 1402 can send raw sensor data to the computing device 1442. In other examples, the sensorpod platform 1402 can send processed sensor data and / or representations of sensor data to the computing device 1442. In some examples, the sensorpod platform 1402 can send sensor data to the computing device 1442 at a specific frequency, after a predetermined time period, or in near real-time. In some cases, the sensorpod platform 1402 can send sensor data (raw or processed) to the computing device 1442 as one or more log files.

[0141] The computing device 1442 can receive sensor data (raw or processed) and perform calibration operations on the data. In at least one example, the computing device 1442 may include one or more processors 1444 and a memory 1446 communicatively coupled to the processors 1444. In the illustrated example, the memory 1446 of the computing device 1442 stores a calibration component 1448, a semantic segmentation component 1450, and / or a map 1452. The calibration component 1448 may include functionality for performing operations similar to those described above in the context of the calibration component 1428, the semantic segmentation component 1450 may include functionality for performing operations similar to those described above in the context of the semantic segmentation component 1430, and the map 1452 may correspond to the map 1438. In some examples, processor 1444 and memory 1446 may include similar functions and / or structures to those described above with respect to processor 1416 and memory 1418.

[0142] Exemplary cleaning systems and techniques Figure 15 shows a diagram of the cleaning system 1500. In this example, the cleaning system 1500 may include a first nozzle 1502 and a second nozzle 1504 positioned adjacent to and facing the first sensing surface 1506 of the first sensor 1508.

[0143] In the example, the nozzles are configured to direct fluid onto the sensing surface to clean it. For example, the first nozzle 1502 applies the first fluid 1510 onto the first sensing surface 1506. In the example, the first fluid 1510 is applied at an angle to the sensing surface. In the example, the second nozzle 1504 is configured to apply another fluid in a similar manner to clean the sensing surface.

[0144] In the example, the various sensors in the sensor pod have two nozzles directed towards the sensing surface to apply a cleaning fluid. In the example, the cleaning fluid may be the same fluid or may be different. For example, the cleaning fluid may be compressed air, water, detergent, de-icing agent, alcohol, or other liquid agent. In the example, some sensors in the sensor pod may have various nozzles directed towards the sensing surface, where the cleaning fluid is the same. In the example, some sensors in the sensor pod may have various nozzles directed towards the sensing surface, where each nozzle applies a different cleaning fluid.

[0145] In the example, the cleaning system 1500 includes a first manifold and / or reservoir for distributing the cleaning fluid to the nozzles. For example, Figure 1 shows a cleaning system 134 which includes a reservoir, a manifold, fluid, pressurizer, pump, and valves, and is controlled by a controller 132 to apply the fluid to the sensing surface through one or more nozzles of a sensor pod. In the example, the cleaning system 134 may have various manifolds and reservoirs to control the delivery of various fluids.

[0146] In the example, the nozzle is configured to apply a fluid, for example, a first fluid 1510, at an angle over the sensing surface. For example, the angle may be between 40 and 85 degrees from the perpendicular (normal) direction of the sensing surface. In the example, this allows for more efficient and effective cleaning of the sensing surface. In the example, this installation allows the nozzle to be located outside the sensing field of view of the sensor or to cover only a small portion of the field of view. In the example, the nozzle is positioned adjacent to the sensing surface but outside at least a portion of the sensing field of view. In the example, the nozzle is positioned adjacent to the sensing surface but covers at least a portion of the sensing field of view. In the example, at least a portion of the sensing field of view that is covered by the nozzle is also covered by a portion of the sensor pod other than the nozzle. For example, a portion of the sensor field of view may already be blocked by a portion of the sensor pod, and the nozzle may be positioned in the effective shade of the blocked portion of the sensor pod.

[0147] In the examples, the sensing surface may vary depending on the sensor. For example, the sensing surface may include a lens, shield, windscreen, casing, or radome. In the examples, the sensing surface may be an integral part of the associated sensor or a part separate from the associated sensor, and may include any material and / or medium on which the associated sensor can receive signals indicating the environment, and therefore may include, but are not limited to, transparent materials, translucent materials, glass, polymers, polycarbonates, and / or combinations thereof. In examples involving sensor types such as RADAR, SONAR, and / or other non-optical sensor types, the sensing surface may or may not transmit signals transmitted and / or received by the sensor.

[0148] In the example, the sensor pod also includes a heater. In the example, each sensor has a heater. In the example, the heater is an electrophotosensitive heater that converts electrical energy into heat. In the example, the heater is coupled between the sensor and the frame on which the sensor is mounted. In the example, the heater is coupled to the cleaning system 1500. In the example, the heater associated with the sensor may be operated to reduce, remove, or facilitate cleaning the sensor or maintaining a clean surface. For example, the heater may be used to reduce, remove, or facilitate the removal of frost or ice that may affect the associated sensor. In the example, the heater may be used in conjunction with one or more of the fluids supplied by the nozzle. For example, the heater may be activated to increase the temperature of the sensor before the use of the fluid. In the example, the heater may preheat the sensor to reduce any cooling caused by subsequent application of the fluid. In the example, the heater may increase the temperature of the sensor or sensing surface so that any residual fluid or other environmental sources supplied by the nozzle evaporate or flash off from the sensor and / or sensing surface.

[0149] Figure 16 is a flowchart illustrating an exemplary method and process involving a vehicle having a sensor pod with a cleaning system. For convenience and ease of understanding, the method shown in Figure 12 is described in relation to one or more of the vehicles, sensor pods, and / or systems shown in Figures 1 to 15. However, the method and process shown in Figure 16 is not limited to being carried out using the vehicles, sensor pods, sensors, and / or systems shown in Figures 1 to 15, but may be implemented using any of the other vehicles, sensor pods, and / or systems described herein, as well as vehicles, sensor pods, and / or systems not described herein. Furthermore, the vehicles, sensor pods, and / or systems described herein are not limited to carrying out the method and process shown in Figure 16.

[0150] Figure 16 shows an exemplary process 1600 for cleaning sensors in a vehicle sensor pod. In operation 1602, in response to a first trigger, the cleaning system applies a first pressurized fluid to a first nozzle on the sensor pod. In this example, the first trigger may include an indication that the sensors in the sensor pod require cleaning, an indication that a threshold time has elapsed since the last cleaning of the sensors, or a weather forecast indicating a weather event or the absence of a weather event.

[0151] In some examples, the system may use one or more of the techniques described in U.S. Patent Application No. 15 / 837,953 filed December 11, 2017, U.S. Patent Application No. 15 / 944,240 filed April 3, 2018, and / or U.S. Patent Application No. 16 / 011,335 filed June 18, 2018, whose disclosures are incorporated herein by reference, to determine that there is a fault on the sensor.

[0152] In operation 1604, the first nozzle directs the first pressurized fluid onto the sensing surface of the sensor.

[0153] In operation 1606, in response to a second trigger, the cleaning system ceases supplying the first pressurized fluid to the first nozzle. In this example, the second trigger may include an indication that the sensing surface is clean or an indication that a threshold application time has elapsed since the first pressurized fluid began to be directed onto the sensing surface.

[0154] In operation 1608, in response to a third trigger, the cleaning system may supply a second pressurized fluid to a second nozzle on the sensor pod. In the example, the second nozzle may be adjacent to the sensor on the sensor pod. In the example, the second pressurized fluid is different from the first pressurized fluid. In the example, the second pressurized fluid is the same as the first pressurized fluid. In the example, the third trigger may include an indication that the sensor on the sensor pod requires cleaning, an indication that another threshold time has elapsed since the previous cleaning, an indication that the first nozzle has stopped supplying the first pressurized fluid, an indication that liquid is on the sensing surface, an indication that the weather forecast indicates a meteorological event, or an indication that the weather forecast indicates the absence of a meteorological event. For example, when the first pressurized fluid is a liquid, a liquid residue may remain on the sensing surface. In the example, when the second pressurized fluid is compressed air, the second nozzle may apply compressed air to blow away any residual liquid from the sensing surface. In the example, when the second pressurized fluid is compressed air, the air can be used to remove rain or other liquids from the sensing surface.

[0155] In operation 1610, in response to a fourth trigger, the cleaning system ceases supplying the second pressurized fluid to the second nozzle. In this example, the fourth trigger may include an indication that the sensing surface is clean or that a threshold application time has elapsed since the second pressurized fluid began to be directed onto the sensing surface. The cleaning system may return to operation 1602 once the first trigger is experienced.

[0156] In the examples, the cleaning system may supply pressurized fluid from a central reservoir. In these examples, the cleaning system may alternate between sensors on one pod before alternating between sensors on another pod. In the examples, sensors facing the direction of travel may be preferred over sensors facing away from the direction of travel.

[0157] Exemplary clause Any of the exemplary provisions in this section may be used in conjunction with any other exemplary provisions and / or any other examples of embodiments described herein.

[0158] A: A sensor pod comprising: a frame having a mounting interface for detachably coupling the sensor pod to a vehicle; a first sensor detachably mounted at a first position on the frame, the first sensor having a first field of view; a second sensor mounted at a second position on the frame, having a second field of view, the second sensor having at least a portion of the first field of view overlapping with at least a portion of the second field of view; and a third sensor mounted at a third position on the frame, having a third field of view, the first sensor and the second sensor having at least a portion of the first field of view overlapping with at least a portion of the second field of view, the first sensor and the second sensor being of a first type, and the third sensor being of a second type different from the first type.

[0159] B: The sensor pod described in paragraph A, wherein the first type of sensor comprises an imaging sensor, and the second type of sensor comprises a LiDAR sensor.

[0160] C: The sensor pod according to paragraph A or B, further comprising: a fourth sensor detachably mounted at a fourth position on the frame, the fourth sensor having a fourth field of view, wherein at least a portion of the second field of view and a portion of the third field of view overlap with at least a portion of the fourth field of view; and a fifth sensor mounted at a fifth position on the frame, having a fifth field of view, wherein at least a portion of the fourth field of view overlaps with at least a portion of the fifth field of view.

[0161] D: A sensor pod as described in any one of paragraphs A to C, wherein the first sensor comprises a first camera whose optical axis is oriented in a first direction, the second sensor comprises a second camera whose optical axis is oriented in a second direction, and the fourth sensor comprises a third camera whose optical axis is oriented in a third direction substantially opposite to the first direction, the second direction being between the first and third directions, and the first field of view and the second field of view being different from the third field of view.

[0162] E: The sensor pod according to any one paragraph A to D, comprising a cast frame, the cast frame having a first cast surface at a first position configured to support a first sensor, a second cast surface at a second position configured to support a second sensor, and a third cast surface at a third position configured to support a third sensor.

[0163] F: The sensor pod according to any one paragraph A to E, further comprising a cast frame with a cast mounting surface at a sixth position configured for mounting the sensor pod to a vehicle.

[0164] G: A cleaning system for cleaning one or more sensors, the cleaning system further comprising a sensor pod according to any one paragraph A to F, wherein the cleaning system comprises a variety of types of fluids applicable to at least one sensor of the sensor pod.

[0165] H: A supply harness that can be coupled to a vehicle, the sensor pod according to any one paragraph A to G, further comprising a supply harness having a liquid connection and a pressurized air connection for supplying fluid and pressurized air to a cleaning system and a power connection for supplying power to one or more sensors.

[0166] I: A sensor pod according to any one paragraph A to H, further comprising a sensor harness electrically coupled to the first, second, third, fourth, and fifth sensors for transmitting sensor data from the first, second, third, fourth, and fifth sensors to the vehicle's computing system, which can be electrically coupled to the vehicle.

[0167] J: The sensor pod is asymmetrical, and the mounting interface is located in a sixth position substantially opposite to the second position, and the mount of the sensor pod coupled to the vehicle protrudes from the side of the sensor pod as described in any one paragraph of paragraphs A to I.

[0168] K: A vehicle comprising a main body having four quarter sections, and two sensor pods, the sensor pods of the two sensor pods being positioned in different quarter sections of the four quarter sections of the vehicle and coupled to the vehicle via mounting arms, the sensor pods of the two sensor pods comprising a frame having a mounting interface detachably coupled to the mounting arms, a first sensor detachably mounted at a first position on the frame, the first sensor having a first field of view, and a second sensor mounted at a second position on the frame, having a second field of view, the first field of view being at least as large as the second field of view A vehicle comprising two sensor pods, each comprising: a second sensor having a second position which partially overlaps with the first, with the second position located substantially opposite the mounting arm; a third sensor having a third field of view, mounted at a third position on the frame, with at least a portion of the second field of view overlapping with at least a portion of the third field of view; and a fourth sensor having a fourth field of view, with at least a portion of the first, second, and third field of view overlapping with at least a portion of the fourth field of view.

[0169] L: The vehicle described in paragraph K, wherein the first sensor, the second sensor, and the third sensor are of the first type, and the fourth sensor is of the second type.

[0170] M: The vehicle according to paragraph K or L, wherein the main body has a first side, a second side, a first vertical end, and a second vertical end, the first sensor pod is disposed in a first quarter section of the vehicle located on the first side and the first vertical end, the second sensor pod is disposed in a second quarter section of the vehicle located on the second side and the first vertical end, the third sensor pod is disposed in a third quarter section of the vehicle located on the first side and the second vertical end, and the fourth sensor pod is disposed in a fourth quarter section of the vehicle located on the second side and the second vertical end, the first sensor pod and the fourth sensor pod have a first shape, and the second sensor pod and the third sensor pod have a second shape which is a mirror image of the first shape.

[0171] N: A system comprising a frame having a mounting interface for detachably coupling the system to a mounting arm of a vehicle; a first sensor detachably mounted at a first position on the frame, the first sensor having a first field of view; a second sensor mounted at a second position on the frame, having a second field of view, the second sensor having at least a portion of the first field of view overlapping with at least a portion of the second field of view, and the second position being substantially opposite the mounting arm; and a third sensor mounted at a third position on the frame, having a third field of view, the second sensor having at least a portion of the second field of view overlapping with at least a portion of the third field of view, the first and second sensors being of a first type, and the third sensor being of a second type different from the first type.

[0172] O: The system according to paragraph N, wherein the first type of sensor comprises an imaging sensor, and the second type of sensor comprises a LiDAR sensor.

[0173] P: The system according to paragraph N or O, further comprising a fourth sensor detachably mounted at a fourth position on the frame, the fourth sensor having a fourth field of view, the second field of view overlapping with at least a portion of the fourth field of view.

[0174] Q: The system described in any one paragraph N to P, wherein the first sensor comprises a first camera whose optical axis is oriented in a first direction, the second sensor comprises a second camera whose optical axis is oriented in a second direction, and the fourth sensor comprises a third camera whose optical axis is oriented in a third direction substantially opposite to the first direction, the second direction being between the first and third directions, and the first field of view and the second field of view being different from the third field of view.

[0175] R: The system according to any one paragraph of paragraphs N to Q, wherein the frame comprises a cast frame, the cast frame comprising a first cast surface at a first position configured to support a first sensor, a second cast surface at a second position configured to support a second sensor, a third cast surface at a third position configured to support a third sensor, and a sixth cast mounting surface configured to mount the system to a vehicle, the first sensor being mounted below a first tolerance threshold relative to the second sensor, and the second sensor being mounted below a second tolerance threshold relative to the vehicle.

[0176] S: A cleaning system for cleaning one or more sensors, the cleaning system further comprising a cleaning system comprising a variety of types of fluids applicable to at least one sensor, as described in any one paragraph N to R.

[0177] T: A supply harness that can be coupled to a vehicle, the supply harness further comprising a liquid connection and a pressurized air connection for supplying fluid and pressurized air to a cleaning system and a power connection for supplying power to one or more sensors, according to any one paragraph of paragraphs N to S.

[0178] U: A vehicle comprising a main body having a first end along the longitudinal axis and a second end distal to the first end, and a first sensor pod comprising various sensors disposed in a first sensor pod housing, wherein the first sensor pod is detachably coupled to the main body at a first position, the first position being at an altitude above the ground interface and adjacent to the first end which is spaced apart from the longitudinal axis in a first direction along the first transverse axis, and a second sensor pod comprising various sensors disposed in a second sensor pod housing, the second sensor pod The system comprises a second sensor pod detachably coupled to the main body in a second position, the second position being adjacent to a first end that, at its height, is spaced apart from the vertical axis in a second direction opposite to the first direction along the first transverse axis, and a third sensor pod comprising various sensors disposed within a third sensor pod housing, the third sensor pod detachably coupled to the main body in a third position, the third position being adjacent to a second end that, at its height, is spaced apart from the vertical axis in a first direction along the second transverse axis, and a fourth sensor pod housing A fourth sensor pod equipped with various sensors arranged thereon, the fourth sensor pod being detachably coupled to the main body at a fourth position, the fourth position being adjacent to a second end located at an altitude adjacent to the second end which is spaced apart from the vertical axis in a second direction opposite to the first direction along the second horizontal axis, the first sensor pod having a first effective sensor field of view, the second sensor pod having a second effective sensor field of view, the third sensor pod having a third effective sensor field of view, the fourth sensor pod having a fourth effective sensor field of view, and the first effective sensor pod having a first effective sensor field of view A vehicle in which the sensor field of view overlaps with at least a portion of the second effective sensor field of view and at least a portion of the third effective sensor field of view, the second effective sensor field of view overlaps with at least a portion of the first effective sensor field of view and at least a portion of the fourth effective sensor field of view, the third effective sensor field of view overlaps with at least a portion of the first effective sensor field of view and at least a portion of the fourth effective sensor field of view, and the fourth effective sensor field of view overlaps with at least a portion of the second effective sensor field of view and at least a portion of the third effective sensor field of view.

[0179] V: The vehicle according to paragraph U, wherein the first effective sensor field of the first sensor pod overlaps with at least a portion of the fourth effective sensor field of the fourth sensor pod.

[0180] W: The vehicle described in paragraph U or V, wherein the field of view of the first effective sensor field of view, the second effective sensor field of view, the third effective sensor field of view, and the fourth effective sensor field of view is between 260 degrees and 280 degrees.

[0181] X: The first effective sensor field of view, the second effective sensor field of view, the third effective sensor field of view, and the fourth effective sensor field of view are the vehicle described in any one paragraph of paragraphs U to W that collectively provides a 360-degree effective vehicle field of view at a first distance from the vehicle.

[0182] Y: The first sensor pod protrudes at least 8 inches from the first end of the main body and at least 8 inches from the side of the vehicle as described in any one paragraph U to X.

[0183] Z: A vehicle according to any one paragraph U to Y, wherein at least three of the first effective sensor field of view, the second effective sensor field of view, the third effective sensor field of view, or the fourth effective sensor field of view collectively provide a 360-degree effective vehicle field of view at a first distance from the vehicle.

[0184] AA: A vehicle according to any one paragraph U through Z, wherein one or more sensor pods protrude at a distance of at least 8 inches from the longitudinal end of the vehicle body and at a distance of at least 8 inches from the side.

[0185] BB: Vehicles described in any paragraph U through AA, where the altitude is 4 feet above and 1 foot above the top of the vehicle.

[0186] CC: The vehicle described in any one paragraph U to BB, wherein the first sensor pod and the third sensor pod are equipped with a first type sensor pod, and the second sensor pod and the fourth sensor pod are equipped with a second type sensor pod, the second type sensor pod being a mirror image of the first type sensor pod.

[0187] DD: A vehicle according to any one paragraph U to CC, further comprising another sensor disposed on a separate body from the first sensor pod, the second sensor pod, the third sensor pod, and the fourth sensor pod, the other sensor having a fifth field of view that overlaps with at least a portion of two of the first effective sensor field of view, the second effective sensor field of view, the third effective sensor field of view, or the fourth effective sensor field of view.

[0188] EE: A vehicle as described in any one paragraph U to DD, wherein the first position is within a first distance from the first end, and the distance of the first position from the vertical axis is greater than a second distance from the vertical axis such that the effective sensor field of view of the first sensor pod includes a view of the vehicle and of an object behind an obstacle located at the first position spaced away from the vehicle.

[0189] FF: Sensor system comprising a first sensor pod comprising various sensors disposed in a first sensor pod housing, the first sensor pod being removablely connectable to the vehicle body at a first position, the first position being at an altitude above the ground interface and adjacent to a first end spaced apart from the longitudinal axis of the body in a first direction along a first transverse axis, and a second sensor pod comprising various sensors disposed in a second sensor pod housing, the second sensor pod being removable from the body at a second position A third sensor pod comprising a second sensor pod that is removably connectable to the main body at a third position, the third position being adjacent to the first end which is spaced at its height in a second direction opposite to the first direction along the first transverse axis and away from the longitudinal axis, and the second end being distal to the first end along the longitudinal axis. The fourth sensor pod comprises a variety of sensors disposed within the fourth sensor pod housing, the fourth sensor pod being detachably coupled to the main body at a fourth position, the fourth position being adjacent to a second end that, at its height, is spaced apart from the vertical axis in a second direction opposite to the first direction along the second transverse axis, the first sensor pod having a first effective sensor field of view, the second sensor pod having a second effective sensor field of view, the third sensor pod having a third effective sensor field of view, and the fourth sensor pod It has a fourth effective sensor field of view, the first effective sensor field of view overlaps with at least a portion of the second effective sensor field of view and at least a portion of the third effective sensor field of view, the second effective sensor field of view overlaps with at least a portion of the first effective sensor field of view and at least a portion of the fourth effective sensor field of view, the third effective sensor field of view overlaps with at least a portion of the first effective sensor field of view and at least a portion of the fourth effective sensor field of view, the fourth effective sensor field of view overlaps with at least a portion of the second effective sensor field of view,A sensor system that overlaps with at least a portion of the third effective sensor field of view.

[0190] GG: The sensor system according to paragraph FF, wherein the first effective sensor field of view of the first sensor pod overlaps with at least a portion of the fourth effective sensor field of view of the fourth sensor pod at a first distance from the vehicle.

[0191] HH: The sensor system according to paragraph FF or GG, wherein the first effective sensor field of view, the second effective sensor field of view, the third effective sensor field of view, and the fourth effective sensor field of view collectively provide a 360-degree effective vehicle field of view at a first distance from the vehicle.

[0192] II: A sensor system according to any one paragraph of paragraphs FF to HH, wherein the first effective sensor field of view, the second effective sensor field of view, the third effective sensor field of view, and the fourth effective sensor field of view are at least 270 degrees.

[0193] JJ: A sensor system according to any one paragraph FF to II, wherein at least three of the first effective sensor field of view, the second effective sensor field of view, the third effective sensor field of view, or the fourth effective sensor field of view collectively provide a 360-degree effective vehicle field of view at a first distance from the vehicle.

[0194] KK: A sensor system according to any one paragraph FF to JJ, wherein one or more sensor pods protrude at a distance of at least 8 inches from the longitudinal end of the vehicle body and at a distance of at least 8 inches from the side.

[0195] LL: A sensor system according to any one paragraph from paragraph FF to KK, wherein the first and third sensor pods are each of the first type of sensor pod, and the second and fourth sensor pods are each of the second type of sensor pod, the second type of sensor pod being a mirror image of the first type of sensor pod.

[0196] MM: A sensor system according to any one paragraph FF to LL, further comprising another sensor disposed on a separate body from the first sensor pod, the second sensor pod, the third sensor pod, and the fourth sensor pod, the other sensor having a fifth field of view that overlaps with at least two portions of the first effective sensor field of view, the second effective sensor field of view, the third effective sensor field of view, or the fourth effective sensor field of view.

[0197] NN: A vehicle comprising a body having a first longitudinal end along a longitudinal axis and a second longitudinal end distal to the first longitudinal end, and a sensor pod detachably coupled to the body, wherein the sensor pod protrudes at a distance of at least 8 inches from the longitudinal end of the vehicle body, protrudes at a distance of at least 8 inches from the side, and is at an altitude above the ground interface, the altitude being 5 feet above and within 2 feet above the body, and the sensor pod comprises a first sensor of a first type and a second sensor of a second type.

[0198] OO: The sensor pod is a vehicle described in paragraph NN having an effective field of view larger than 270 degrees.

[0199] PP: Sensor pod comprising a frame that can be coupled to a vehicle, a sensor coupled to the frame, a collision structure coupled to the frame, having an outer surface configured to interface with a pedestrian during a collision, with at least a portion of the outer surface disposed outside the sensor relative to the frame, and a collision energy absorbing structure disposed between the outer surface and the frame and configured to absorb a portion of the energy transmitted from the collision through the outer surface.

[0200] QQ: The sensor pod described in paragraph PP, wherein the collision energy absorption structure comprises a stress concentration zone during the collision, and the stress concentration zone is configured to cause local plastic deformation of the collision energy absorption structure to absorb energy from the collision.

[0201] RR: A sensor pod according to paragraph PP or QQ, configured such that a stress concentration zone produces local plastic deformation that exceeds a first impact force threshold and falls below a second impact force threshold.

[0202] SS: A sensor pod described in any one paragraph from paragraph PP to RR, whose outer surface is configured to move substantially uniformly relative to the frame during a collision.

[0203] TT: A sensor pod according to any one paragraph from paragraph PP to SS, having an outer surface thickness configured to plastically deform above a first impact force threshold and below a second impact force threshold during a collision.

[0204] UU: A sensor pod described in any one paragraph from paragraph PP to TT, whose outer surface is substantially convex.

[0205] VV: A sensor pod described in any one paragraph from paragraph PP to UU, wherein the outer surface is substantially convex in two orthogonal directions.

[0206] WW: A sensor pod according to any one paragraph from paragraph PP to VV, further comprising another sensor disposed on top of the frame and coupled to the frame by a deformable fastener.

[0207] XX: A sensor pod according to any one paragraph from paragraph PP to WW, comprising a deformable fastener having a stress concentration zone for causing local plastic deformation of the deformable fastener to absorb energy from a collision, and a deformable fastener configured to release other sensors when the collision exceeds a third collision force threshold.

[0208] YY: A sensor pod according to any one paragraph from paragraph PP to XX, further comprising a baffle structure disposed between the sensor and the outer surface, the baffle structure having a crumple zone configured to plastically deform during a collision and absorb energy from the collision.

[0209] ZZ: A system comprising a frame that can be coupled to a vehicle, a sensor coupled to the frame, a collision structure coupled to the frame, having an outer surface configured to interface with a pedestrian during a collision, with at least a portion of the outer surface disposed outside the sensor relative to the frame, and a collision energy absorbing structure disposed between the outer surface and the frame and configured to absorb a portion of the energy transmitted from the collision through the outer surface.

[0210] AAA: The collision energy absorption structure is configured to absorb energy from the collision by providing a stress concentration zone during the collision, the stress concentration zone being configured to cause local plastic deformation of the collision energy absorption structure. (Paragraph ZZ)

[0211] BBB: The system described in paragraph ZZ or AAA, wherein the stress concentration zone is configured to produce local plastic deformation that is above a first impact force threshold and below a second impact force threshold.

[0212] CCC: The outer surface is configured to move substantially uniformly relative to the frame during a collision, as described in any one of paragraphs ZZ through BBB.

[0213] DDD: A system described in any one paragraph from paragraph ZZ to CCC, wherein the outer surface is substantially convex in two orthogonal directions.

[0214] EEE: A vehicle comprising a vehicle body, a sensor pod protruding from the vehicle body, a frame connectable to the vehicle, a sensor connected to the frame, a collision structure connected to the frame, an outer surface configured to interface with a pedestrian during a collision, at least a portion of which is disposed outside the sensor relative to the frame, and a collision energy absorbing structure disposed between the outer surface and the frame, configured to absorb a portion of the energy transmitted through the outer surface from the collision.

[0215] FFF: The vehicle according to paragraph EEE, further comprising another sensor pod protruding from the vehicle body, the sensor pod protruding from a first quarter section of the vehicle body, and the other sensor pod protruding from a second quarter section of the vehicle body.

[0216] GGG: The sensor pod is a first sensor pod protruding from a first quarter section of the vehicle body, and the second sensor pod is a vehicle as described in paragraph EEE or FFF, protruding from the vehicle body in a second quarter section of the vehicle body.

[0217] HHH: A vehicle according to any one paragraph from paragraph EEE to GGG, wherein the collision energy absorption structure comprises a stress concentration zone during a collision, and the stress concentration zone is configured to cause local plastic deformation of the collision energy absorption structure to absorb energy from the collision.

[0218] III: A vehicle according to any one paragraph from paragraph EEE to HHH, wherein the stress concentration zone is configured to produce local plastic deformation that exceeds a first impact force threshold and falls below a second impact force threshold.

[0219] JJJ: A system comprising a vehicle mounting interface, a frame having a first mounting position and a second mounting position, a first sensor coupled to the frame at the first mounting position, a second sensor coupled to the frame at the second mounting position, one or more processors, and a non-temporary computer-readable medium which, when executed, enables one or more processors to receive first sensor data of the system environment from the first sensor, receive second sensor data of the system environment from the second sensor, segment the first sensor data as first segmented data based at least partially on a representation of a first immutable object in the first sensor data, and segment the second sensor data as second segmented data based at least partially on a representation of a second immutable object in the second sensor data. A system comprising a non-temporary computer-readable medium that stores one or more instructions causing the system to perform an action comprising: transforming; determining a subset of first sensor data relating to a first immutable object based at least in part on first segmented data; determining a subset of second sensor data relating to a second immutable object based at least in part on second segmented data; determining a first calibration error relating to a first sensor based at least in part on the subset of first sensor data; determining a second calibration error relating to a second sensor based at least in part on the subset of second sensor data; and determining a system calibration factor relating to the system based at least in part on the first calibration error relating to the first sensor and the second calibration error relating to the second sensor.

[0220] KKK: A system described in paragraph JJJ, comprising a mount coupled to a frame in a vehicle mounting interface, the mount further comprising a mount fixed to the environment.

[0221] LLL: The environment is a system described in paragraph JJJ or KKK that includes the first invariant within a 90-degree field of view.

[0222] MMM: The mount has an indexed orientation, and the first sensor data is received at the first index, and the second sensor data is received at the second index, and the determination is made based at least partially on the orientation of the first index and the orientation of the second index, as described in any one paragraph of paragraphs JJJ to LLL.

[0223] NNN: The system according to any one paragraph of paragraphs JJJ to MMM, further comprising a third sensor coupled to a frame at a third mounting position, wherein the act comprises receiving third sensor data of the system environment from the third sensor; segmenting the third sensor data as third segmented data, at least in part on a representation of a third invariant object in the third sensor data; determining a subset of the third sensor data relating to the third invariant object, at least in part on the third segmented data; determining a third calibration error relating to the third sensor, at least in part on the subset of the third sensor data; and determining a system calibration factor relating to the system, at least in part on the third calibration error relating to the third sensor.

[0224] OOO: The system according to any one paragraph of paragraphs JJJ to NNN, wherein the mount has an indexed orientation, and the first sensor data and the second sensor data are received with the frame oriented to the first index, and the third sensor data are received with the frame oriented to the second index, and the act further comprises receiving the fourth sensor data from the second sensor with the frame at the second index, and determining a system calibration factor related to the system that is at least partially based on the third sensor data and the fourth sensor data.

[0225] PPP: The system described in any paragraph from paragraph JJJ to OOO, where the first invariant is the same as the second invariant.

[0226] QQQ: The first invariant is a system described in any one paragraph JJJ to PPP, comprising a standard, a part of a building, a wall, a corner, a pole, a window, a roof / ceiling line, an intersection of a wall and a ceiling / roof, or a combination thereof.

[0227] RRR: The system described in any paragraph from paragraph JJJ to QQQ, wherein the first sensor is a different type of sensor from the second sensor.

[0228] SSS: The system described in any one paragraph from paragraph JJJ to RRR, wherein the first sensor is a camera and the second sensor is a LiDAR.

[0229] The system described in any one paragraph of paragraphs JJJ to SSS further comprises: receiving sensorpod data representing a third invariant from a sensorpod; determining a sensorpod calibration factor at least in part based on the sensorpod data; and determining, at least in part based on a system calibration factor and a sensorpod calibration factor, whether the sensorpod is compatible with a first sensor and a second sensor mounted on a frame when the sensorpod and frame are mounted on a vehicle.

[0230] UUU: A vehicle comprising a sensor pod having a first sensor and a second sensor, one or more processors, and a non-temporary computer-readable medium storing one or more instructions that, when executed, cause one or more processors to perform an action comprising: receiving sensor data of the vehicle environment from the first sensor; segmenting the sensor data as segmented data, at least in part on representations of immutable objects in the sensor data; determining a subset of sensor data related to immutable objects, at least in part on the segmented data; determining a calibration error related to the first sensor, at least in part on the subset of sensor data; and determining an estimated calibration error related to the second sensor, at least in part on the calibration error and a sensor pod calibration factor.

[0231] VVV: The sensor pod calibration factor is a vehicle as described in paragraph UUU, which is at least partially based on a first calibration error related to the first sensor, which represents a first invariant, and a second calibration error related to the second sensor, which represents a second invariant, which is at least partially based on second data received from the second sensor.

[0232] WWW: The first invariant is the same as the second invariant for the vehicle described in paragraph UUU or VVV.

[0233] XXX: A vehicle described in any one paragraph from paragraph UUU to WWW, wherein the first sensor is a camera and the second sensor is a LiDAR.

[0234] YYY: The vehicle described in any one paragraph of paragraphs UUU to XXX, further comprising: receiving sensor pod data representing a third invariant from another sensor pod; determining another sensor pod calibration factor at least in part based on the sensor pod data; and determining, at least in part based on the sensor pod calibration factor and the other sensor pod calibration factor, whether the sensor pod and the other sensor pod have a suitable effective field of view when they are mounted on the vehicle.

[0235] ZZZ: A method comprising: receiving first sensor data representing a first invariant from a first sensor pod; determining a first calibration factor based at least in part on the first sensor data; receiving second sensor data representing a second invariant from a second sensor pod; determining a second calibration factor based at least in part on the second sensor data; and determining whether the first sensor pod and the second sensor pod have a suitable effective field of view based at least in part on the first calibration factor and the second calibration factor.

[0236] AAAA: The method of paragraph ZZZ, further comprising the step of pairing a first sensor pod with a second sensor pod for installation on a vehicle when it is determined that it has a suitable effective field of view.

[0237] The method of paragraph ZZZ or AAAA, further comprising the step of installing a first sensor pod on a vehicle on which a second sensor pod is already installed, when it is determined that it has a suitable field of view.

[0238] CCCC: The method of the first sensor data comprising any one paragraph from paragraph ZZZ to BBBB, wherein the first sensor data comprises data from multiple sensors of the first sensor pod.

[0239] DDDD: A sensor pod comprising: a first sensor coupled to a frame, the first sensor having a first sensing surface and a sensing field of view; a first nozzle disposed adjacent to and directed toward the first sensing surface, configured to clean the first sensing surface by applying a first fluid to it; a second nozzle disposed adjacent to and directed toward the first sensing surface, configured to clean the first sensing surface by applying a second fluid to it; and a second sensor coupled to a frame, the second sensor having a second A sensor pod comprising: a second sensor having a detection surface and a detection field of view; a third nozzle disposed adjacent to and directed toward the second detection surface, the third nozzle configured to apply a first fluid to the second detection surface to clean the second detection surface; a fourth nozzle disposed adjacent to and directed toward the second detection surface, the fourth nozzle configured to apply a second fluid to the second detection surface to clean the second detection surface; a first manifold for distributing the first fluid to the first nozzle and the third nozzle; and a second manifold for distributing the second fluid to the second nozzle and the fourth nozzle.

[0240] EEEE: A sensor pod as described in paragraph DDDD, comprising a first fluid of air, liquid, water, detergent, cleaning solution, de-icing agent, or alcohol, and a second fluid of air, liquid, water, detergent, cleaning solution, de-icing agent, or alcohol.

[0241] FFFF: The first fluid is a sensor pod described in paragraph DDDD or EEEE, which is different from the second fluid.

[0242] GGGG: A sensor pod according to any one paragraph from paragraph DDDD to FFFF, wherein the first nozzle is configured to apply fluid at an oblique angle to the first sensing surface.

[0243] HHHH: The first fluid is supplied to the first sensing surface from a reservoir that supplies fluid to another sensor pod, according to one of the paragraphs DDDD to GGGG.

[0244] IIII: The nozzle is a sensor pod described in any one of paragraphs DDDD to HHHH, which is adjacent to the detection surface and located outside the detection field of view.

[0245] JJJJ: The nozzle is positioned behind a portion of the sensor pod within the detection field that obstructs the detection field at least partially, as described in any paragraph DDDD to JJJJ.

[0246] KKKK: A sensor pod as described in any one paragraph from paragraph DDDD to JJJJ, further comprising one or more additional nozzles substantially equidistant from the sensor, sensing surface, or field of view.

[0247] LLLL: The sensing surface is a sensor pod as described in any one paragraph from paragraph DDDD to KKKK, comprising a lens, shield, windscreen, casing, or radome.

[0248] MMMM: A supply harness that can be coupled to a vehicle, the supply harness comprising one or more of the following: power, liquid, or pressurized air, further comprising the sensor pod according to any one paragraph of paragraphs DDDD to LLLL.

[0249] NNNN: A method for operating a sensor pod, comprising the steps of: providing pressurized fluid to a nozzle on the sensor pod in response to a first trigger, wherein the nozzle is adjacent to a sensor on the sensor pod; directing the pressurized fluid through the nozzle onto the sensing surface of the sensor; and ceasing to provide pressurized fluid to the nozzle in response to a second trigger.

[0250] The method according to paragraph NNNN, further comprising: receiving first sensor data indicating a first state; generating a first signal indicating a first trigger, at least partially based on the first sensor data, wherein the step of supplying pressurized fluid is at least partially based on the first signal; receiving second sensor data indicating a second state; generating a second signal indicating a second trigger, at least partially based on the second sensor data, wherein the step of ceasing to supply pressurized fluid is at least partially based on the second signal.

[0251] PPPP: The method of paragraph NNNN or OOOO, wherein the first trigger is an indication that the sensor of the sensor pod needs cleaning, an indication that a threshold time has elapsed since the last cleaning, or a weather forecast, one or more of the above.

[0252] QQQQ: The method of any one paragraph from paragraphs NNNN to PPPP, wherein the second trigger is an indication that the sensing surface is clean or that a threshold application time has elapsed since the pressurized fluid began to be directed onto the sensing surface.

[0253] The method of any one paragraph of paragraphs NNNN to QQQQ, further comprising the steps of: providing another pressurized fluid to another nozzle on a sensor pod in response to a third trigger, wherein the other nozzle is adjacent to a sensor on the sensor pod; directing the other pressurized fluid through the other nozzle onto the sensing surface of the sensor; and ceasing to provide the other pressurized fluid to the other nozzle in response to a fourth trigger.

[0254] SSSS: The method of any one paragraph of paragraphs NNNN to RRRR, wherein the third trigger is an indication that the sensor of the sensor pod requires cleaning, an indication that another threshold time has elapsed since the previous cleaning, an indication that the nozzle has stopped supplying pressurized fluid, or an indication that liquid is on the sensing surface.

[0255] TTTT: A fourth trigger comprising an indication that the sensing surface is clean, or an indication that a threshold application time has elapsed since the sensing surface began to be directed onto the sensing surface, according to the method of any one paragraph from NNNN to SSSS.

[0256] UUUU: The method of any one paragraph of paragraphs NNNN to TTTT, wherein the first trigger comprises an indication that the sensor of the sensor pod requires cleaning or that a threshold time has elapsed since the previous cleaning; the second trigger comprises an indication that the sensing surface is clean or that a threshold application time has elapsed since pressurized fluid began to be directed onto the sensing surface; the third trigger comprises an indication that the sensor of the sensor pod requires cleaning, that another threshold time has elapsed since the previous cleaning, that the nozzle has stopped supplying pressurized fluid, or that liquid is on the sensing surface; and the fourth trigger comprises an indication that the sensing surface is clear or that a threshold application time has elapsed since other pressurized fluid began to be directed onto the sensing surface.

[0257] VVVV: The method according to any one of the paragraphs NNNN to UUUU, wherein the pressurized fluid is a liquid and the other pressurized fluid is compressed air.

[0258] WWWW: A system comprising a sensor pod having a variety of sensors, the sensor of the variety of sensors having a sensing surface, a variety of nozzles disposed on the sensor, the first nozzle of the variety of nozzles being configured to apply a first fluid, the second nozzle of the variety of nozzles being configured to apply a second fluid, and a manifold for distributing fluid to the variety of sensors.

[0259] XXXX: The first nozzle is configured to apply the fluid at an oblique angle to the sensing surface in the system described in paragraph WWWW.

[0260] While the exemplary clauses described above are explained in relation to a specific implementation, it should be understood that in the context of this specification, the content of the exemplary clauses may also be implemented through methods, devices, systems, computer-readable media, and / or other implementations.

[0261] conclusion While one or more examples of the techniques described herein have been described, various modifications, additions, substitutions, and equivalents thereof are included within the scope of the techniques described herein.

[0262] In the description of examples, please refer to the attached drawings, which form part of this Spec., illustrating specific examples of the subject matter claimed as examples. Please understand that other examples may be used and that changes or modifications, such as structural changes, may be made. Such examples, changes or modifications do not necessarily constitute a deviation from the intended scope of the claimed subject matter. While the steps in this Spec. may be presented in a certain order, in some cases the order may be changed, and therefore some inputs may be given at different times or in different orders without altering the function of the systems and methods described. The procedures disclosed may also be performed in different orders. Furthermore, the various calculations in this Spec. may not be performed in the order disclosed, and other examples using alternative orders of calculations can be readily implemented. In addition to being reordered, calculations may also be broken down into subcalculations having the same results.

Claims

1. It is a sensor pod, A frame having a mounting interface for detachably connecting the sensor pod to the vehicle, A first sensor that is removably mounted at a first position on the frame, wherein the first sensor has a first field of view, A second sensor having a second field of view, mounted at a second position on the frame, wherein at least a first portion of the first field of view overlaps with at least a first portion of the second field of view. A third sensor having a third field of view, mounted at a third position on the frame, wherein at least a second portion of the first field of view and at least a second portion of the second field of view overlap with at least a first portion of the third field of view, the first sensor and the second sensor are of a first type, and the third sensor is of a second type different from the first type. Equipped with, The frame comprises a cast frame, the cast frame comprising a cast mounting surface configured for mounting the sensor pod to the vehicle at a sixth position substantially opposite to the second position, The sensor pod further comprises a cleaning system for cleaning a plurality of sensors of the sensor pod, the cleaning system comprises a variety of fluids applicable to the plurality of sensors of the sensor pod, the cleaning system comprises nozzles disposed toward one of the plurality of sensors of the sensor pod, and the cleaning system cleans each of the plurality of sensors of the sensor pod by supplying pressurized fluid from the corresponding nozzles.

2. The first type of sensor comprises an imaging sensor, The second type of sensor includes a LIDAR sensor. The sensor pod according to claim 1.

3. A fourth sensor, which is detachably mounted at a fourth position of the frame, the fourth sensor having a fourth field of view, wherein at least a third portion of the second field of view and a second portion of the third field of view overlap with at least a first portion of the fourth field of view. A fifth sensor having a fifth field of view, mounted at a fifth position on the frame, wherein at least a second portion of the fourth field of view overlaps with at least a first portion of the fifth field of view. The sensor pod according to claim 1, further comprising:

4. The sensor pod according to claim 3, wherein the first sensor comprises a first camera whose optical axis is oriented in a first direction, the second sensor comprises a second camera whose optical axis is oriented in a second direction, and the fourth sensor comprises a third camera whose optical axis is oriented in a third direction included in the third field of view, the second direction being between the first direction and the third direction, and the first field of view and the second field of view being different from the third field of view.

5. The aforementioned cast frame is, A first casting surface at the first position configured to support the first sensor, A second casting surface at the second position configured to support the second sensor, The third casting surface at the third position configured to support the third sensor and A sensor pod according to claim 1, comprising:

6. The sensor pod according to claim 1, wherein the cleaning system comprises a variety of fluids applicable to the plurality of sensors of the sensor pod.

7. A sensor pod according to claim 6, further comprising a supply harness that can be coupled to the vehicle, the supply harness comprising a liquid connection and a pressurized air connection for supplying fluid and pressurized air to the cleaning system and a power connection for supplying power to the plurality of sensors.

8. The sensor pod according to claim 1, further comprising a sensor harness electrically coupled to the first sensor, second sensor, third sensor, fourth sensor, and fifth sensor, which is electrically connectable to the vehicle and transmits sensor data from the first sensor, second sensor, third sensor, fourth sensor, and fifth sensor to the vehicle's computing system.

9. The aforementioned sensor pod is asymmetrical, The sensor pod according to claim 1, wherein the mount for the sensor pod coupled to the vehicle is for attaching the sensor pod to the vehicle.

10. It is a system, A frame having a mounting interface that detachably connects the system to the mounting arm of the vehicle, A first sensor that is removably mounted at a first position on the frame, wherein the first sensor has a first field of view, A second sensor having a second field of view, mounted at a second position on the frame, wherein the second position is oriented relative to the first position so that at least a first portion of the first field of view overlaps with at least a first portion of the second field of view. A third sensor having a third field of view, mounted at a third position on the frame, wherein at least a second portion of the second field of view overlaps with at least a first portion of the third field of view, the first sensor and the second sensor are of a first type, and the third sensor is of a second type different from the first type. Equipped with, The frame comprises a cast frame, the cast frame comprising a cast mounting surface configured for mounting a sensor pod to the vehicle at a sixth position substantially opposite to the second position, The sensor pod further comprises a cleaning system for cleaning a plurality of sensors of the sensor pod, the cleaning system comprises a variety of fluids applicable to the plurality of sensors of the sensor pod, the cleaning system comprises nozzles disposed toward one of the plurality of sensors of the sensor pod, and the cleaning system is a system for cleaning each of the plurality of sensors of the sensor pod by supplying pressurized fluid from the corresponding nozzles.

11. The first type of sensor comprises an imaging sensor, The second type of sensor includes a LIDAR sensor. The system according to claim 10.

12. The system according to claim 10, further comprising a fourth sensor detachably mounted at a fourth position of the frame, the fourth sensor having a fourth field of view, wherein at least a third portion of the second field of view overlaps with at least a first portion of the fourth field of view.

13. The system according to claim 12, wherein the first sensor comprises a first camera whose optical axis is oriented in a first direction, the second sensor comprises a second camera whose optical axis is oriented in a second direction, and the fourth sensor comprises a third camera whose optical axis is oriented in a third direction substantially opposite to the first direction, the second direction being between the first and third directions, and the first field of view and the second field of view being different from the third field of view.

14. The aforementioned cast frame is, A first casting surface at the first position configured to support the first sensor, A second casting surface at the second position configured to support the second sensor, The third casting surface at the third position configured to support the third sensor and The system according to claim 10, comprising:

15. The cleaning system according to claim 10, comprising a variety of fluids applicable to the plurality of sensors.

16. The system according to claim 15, further comprising a supply harness connectable to the vehicle, the supply harness comprising a liquid connection and a pressurized air connection for supplying fluid and pressurized air to the cleaning system and a power connection for supplying power to the plurality of sensors.

Images