Collision detection for low energy impacts

Abstract

The present disclosure relates to systems and techniques for identifying collisions, such as relatively low-energy impact collisions involving autonomous vehicles. Sensor data from a first sensor modality in a first array can be used to determine a first estimated collision location, and second sensor data from a second sensor modality in a second array can be used to determine a second estimated collision location. A low-energy collision event can be constituted when the first estimated collision location corresponds to the second estimated collision location.

Description

[Technical field]

[0001] (CROSS REFERENCE TO RELATED APPLICATIONS) This PCT international patent application claims the benefit of priority to U.S. patent application Ser. No. 17 / 245,963, filed April 30, 2021, the disclosure of which is incorporated herein by reference in its entirety. [Background technology]

[0002] In the event of a crash, safety systems on passenger transportation vehicles may be designed to mitigate the impact of the crash, for example, by inflating airbags or tightening physical restraints. In many such safety systems, accelerometers and / or other sensors are installed at the front of the vehicle to detect sudden, large changes in the acceleration or other movement of the vehicle, such as due to a head-on collision with an object, and the detected condition is used to deploy the safety system, for example, by inflating airbags. Conventional safety systems are often not optimized to detect low-energy crashes, where it is not necessary to deploy conventional safety devices to ensure the safety of vehicle occupants. However, low-energy crashes frequently occur during vehicle operation. [Brief description of the drawings]

[0003] The detailed description will be set forth with reference to the accompanying drawings, in which the leftmost digit(s) of a reference number identifies the figure in which the reference number first appears, and the same reference numbers in different drawings indicate similar or identical items.

[0004] [Figure 1] FIG. 1 illustrates an example environment in which the techniques described herein may be implemented. [Diagram 2] FIG. 2 illustrates details of a vehicle equipped with a low-energy impact crash detection system according to an exemplary embodiment of the present disclosure. [Diagram 3] FIG. 3 illustrates a block diagram of an example system for implementing the techniques described herein. [Figure 4]FIG. 4 illustrates an example process for determining a crash event and controlling a vehicle based on the event. [Diagram 5] FIG. 5 illustrates an exemplary process for determining a crash event. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0005] The present disclosure is directed to a vehicle including a low-energy impact crash detection system. The low-energy impact crash detection system is configured to detect a low-energy impact crash and control the vehicle in the event of a low-energy impact crash. The crash detection system may be communicatively coupled to a sensor system associated with the vehicle and configured to receive information from the sensor system. The crash detection system is also operably coupled to a system controller configured to control the vehicle according to actions generated by the crash detection system. In an example, the low-energy impact crash detection system may be added to a conventional or high-energy impact crash detection system used to detect and / or respond to relatively high impact crashes, e.g., to deploy airbags or the like.

[0006] The low-energy impact collision detection system is configured to detect and respond to low-energy impact collisions. For example, a low-energy impact collision may include a medium-speed collision with a low-mass object, such as a collision with a bicycle or scooter, an animal, a pedestrian, or a small object. A low-energy impact collision may also include a low-speed collision with a high-mass object, such as a collision with another vehicle or an obstacle in a parking lot, while the vehicle is stationary or moving slowly.

[0007] Low-energy impact crash events typically do not require deployment of traditional vehicle safety systems such as airbag and air curtain systems or automatic safety devices in automobiles, but may require the vehicle to take some other action, such as, for example, stopping or communicating with an operator. Primary crash detection systems, on the other hand, are associated with traditional vehicle safety systems and are configured to take some action, such as, for example, deploying airbags or tightening safety devices, that are known to help ensure the safety of passengers within the vehicle in the event of a relatively high impact crash.

[0008] An exemplary vehicle, such as an exemplary autonomous vehicle, may include one or more types of sensor modalities for detecting low-energy impact collisions. These sensors may be mounted on the vehicle's body, typically on surfaces exposed to the environment in which the vehicle is operating, or may be embedded in the vehicle's body. In some examples, multiple sensor units may be spatially distributed in different parts of the body configured to provide spatial coverage of the vehicle's body. The density of sensors may also be higher on areas of the body where there is a higher probability of collision with external objects, such as lower parts located at the two ends (e.g., front and rear) of the vehicle. Sensors may also be mounted on or embedded in the wheel covers or fenders of the vehicle. According to some examples, types of sensor modalities may include electromechanical sensors (e.g., impact sensors, inertial sensors) and acoustic sensors (e.g., microphones) for capturing mechanical stresses and sounds / vibrations that may result from a collision, respectively.

[0009] In some examples, the vehicle may be equipped with an electromechanical sensor, which may be an impact sensor in the form of a resistive film strip. Such a sensor detects a change in electrical conductivity when the strip bends. The impact sensor may be mounted on the underside of the vehicle, which may be more likely to be associated with a low-energy impact collision. In some examples, the impact sensor may be incorporated into a body panel of the vehicle in an area that is more likely to be impacted during a low-energy impact collision. For example, the sensors may be located on the ends of the vehicle, the sides of the vehicle, and / or the wheel covers of the vehicle. The impact sensors detect a change in electrical conductivity when the surface to which they are mounted deforms during a low-energy impact collision. In some examples, the electromechanical sensors are arranged in an array.

[0010] The vehicle may also be equipped with an acoustic sensor. For example, the acoustic sensor may include multiple microphones located at spatially separated locations on the vehicle. In some examples, the acoustic sensor may be located in an area of ​​the vehicle that is more likely to collide during a low-energy impact collision. The acoustic sensor may be coupled to a body panel of the vehicle, which may also incorporate an electromechanical sensor. The acoustic sensor may be configured to detect audio associated with the collision. In some examples, by comparing audio received at two audio sensors, the techniques described herein may implement a spatial localization technique to estimate the source of the sound.

[0011] In aspects of the present disclosure, a low energy impact crash detection system may estimate one or more locations of a crash based on sensor data, such as from mechanical and / or acoustic sensors. For example, the techniques described herein may determine an estimate of the crash location relative to the vehicle based on the sensor signals and the known locations of the sensors. For example, in an example crash, the sensor at the crash point registers the highest signal, and the signal strength decreases as the distance from the crash point increases. In general, the strength of the sensor signal depends on the proximity of the crash location to the sensor location. The sensor signal of a non-crash area (e.g., not involved in the crash) may be much lower than the sensor signal in the vicinity of the crash. In one non-limiting example, the sensor signals may be clustered into two groups based on signal strength, for example, by using k-means clustering with k=2. The crash location may be estimated as the spatial center of the sensor cluster with the higher mean.

[0012] Sensor signal levels from non-crash regions may also be learned from training data collected during typical vehicle operation. The learned signal levels corresponding to the non-crash regions may be used as a reference signal. For example, the measured sensor signal may be compared to the reference signal and if the difference is higher than a threshold, the sensor may be identified as being within the crash region. The estimated location of the crash may be calculated as the spatial center of the sensor identified within the crash region, or the estimated location of the crash may correspond to the location of the sensor with the highest signal. Other data may also be considered in the trained model. For example, but not limited to, wind speeds proximate to the vehicle due to vehicle motion and / or current atmospheric conditions may affect the signal generated at the sensor.

[0013] The crash location estimation component may also implement various artificial intelligence (AI) techniques, such as machine learning, to predict the location and type of crash from a set of sensor signals. A training data set corresponding to low-energy impact crashes may be collected from operational data of a vehicle with associated sensors. The training data set may be labeled with the location and type of crash and may include data without crashes. When the trained model is deployed, the crash location estimation component may input the sensor signals to the trained model and receive an estimate of the crash location. In an example, multiple neural networks, each trained with a different type of sensor, may be deployed. Alternatively, a single neural network with aggregated different types of sensor signals as inputs may also be used. The techniques used to estimate the crash location may be similar or different for each type of sensor modality.

[0014] The techniques described herein may also confirm a low-energy impact crash event by checking consistency between crash indications from different types of sensor systems. For example, a first type of sensor (e.g., an electromechanical sensor) may indicate a crash in one area of ​​the vehicle. A second type of sensor (e.g., an acoustic sensor) located in the same area of ​​the vehicle is also expected to indicate a crash in the instance of an actual crash. In the absence of a corresponding signal from the second type of sensor indicating a crash in the same area, the techniques described herein may determine that a low-energy impact crash event has not occurred. In another example, if both the first type of sensor and the second type of sensor indicate a possible crash in the same area of ​​the vehicle, a low-energy impact crash event has occurred.

[0015] In some examples, the perception component may provide additional input to the low-energy impact crash detection system. The perception component may include functionality for performing object detection, segmentation, and / or classification. In some examples, the perception component may output perception data indicating the presence of one or more objects proximate the autonomous vehicle and / or the classification of the object as an object type (e.g., car, pedestrian, cyclist, building, tree, road surface, curb, sidewalk, unknown, etc.). Additionally, the perception component may provide processed sensor data indicating a position and trajectory associated with the detected object and / or a likelihood of a collision with the object based on the vehicle's position and trajectory. In a non-limiting example, the perception data may identify a detected object that may have collided with the vehicle at an estimated collision location determined from the first sensor and / or the second sensor. The presence of such an object may confirm a low-energy impact crash event, and the absence of such an object may indicate that a low-energy impact crash event has not occurred.

[0016] In some examples, the perception component can detect an object proximate to the vehicle and can determine that the object is likely to hit the vehicle. In some examples, the techniques described herein can determine an expected impact location based on the perception data. Information about a possible impact (and its possible impact location) can cause the systems described herein to request signals from other sensors on the vehicle to determine, for example, whether the data suggests or confirms a crash. If either the first type sensor or the second type sensor indicates a crash at the expected impact location, the low-energy impact crash event is confirmed by the event detection component. If both the first and second type sensors do not indicate a crash at the expected location, the event detection component can indicate that a low-energy impact crash has not occurred.

[0017] Implementations of the present disclosure also include acting in response to determining a crash event. For example, the techniques described herein can generate commands for a system controller of a vehicle to control the vehicle based on the detected event. For example, in the case of a low-energy impact crash event, the command generation component can generate commands for the system controller to perform one or more actions, such as stopping the vehicle if currently moving, collecting and transmitting or storing operational data related to the event, and / or initiating communication with a remote operator or other human in or near the vehicle. The techniques described herein also improve safety outcomes. For example, by identifying a low-energy impact, the techniques described herein can identify potential damage to the vehicle and support fixing for the damage, and identify and react to potential damage to objects that may impact the vehicle.

[0018] The techniques, assemblies, and systems described herein may be implemented in a number of ways. Examples are presented below with reference to the figures. Although described in the context of an autonomous vehicle, in some examples, the methods, apparatus, and systems described herein may be applied to a variety of systems. In another example, the methods, apparatus, and systems may be utilized in an aviation or nautical context. Additionally or alternatively, the techniques described herein may be used with real data (e.g., captured using a sensor), simulated data (e.g., generated by a simulator), or any combination of the two.

[0019] FIG. 1 illustrates an example of an autonomous vehicle 100 in which the technology discussed herein may be implemented. The autonomous vehicle may be configured to operate according to the Level 5 classification. The Level 5 classification was issued in 2016 by the National Highway Traffic Safety Administration and describes a vehicle that can perform all safety-critical functions throughout the entire trip without the driver (or passenger) being expected to control the vehicle at all times. In that case, the vehicle may be configured to control all functions from start to finish, including all parking functions, so that the vehicle may not be in human use. This is merely an example, and the systems and methods described herein may be incorporated into any ground vehicle, including traditional automobiles and partially autonomously controlled vehicles, in which all vehicle control is performed by the driver, so that it can be autonomously controlled without driver attention or assistance during certain driving conditions, such as, for example, while operating on limited-access highways. Such partially autonomously controlled vehicles may require driver attention and / or assistance during other driving conditions, such as, for example, while operating on urban streets or during at least some parking functions.

[0020] As shown in FIG. 1, the vehicle 100 has a body 102 that generally defines the surfaces exposed to the exterior of the vehicle, e.g., the environment in which the vehicle 100 is operating. The body 102 may be disposed on, or more generally include, a frame or one or more additional support structures (not shown), as is conventionally known. In the exemplary vehicle 100, the body generally includes a first exterior surface 104 at a first end and a second exterior surface (not visible in FIG. 1) at a second end opposite the first end. The first exterior surface 104 and the second exterior surface are separated along a longitudinal axis X by the length of the vehicle. The body 102 also includes sides or lateral sides 106 (one of which is shown) that extend between the first and second ends of the vehicle 100 and are separated by a vehicle width along a lateral axis Y. The body 102 may further include one or more doors, including, for example, a full or partial door on one or both of the lateral sides 106, to facilitate entry and exit into the interior space of the vehicle 100. The vehicle 100 may also include a windshield and / or other windows, including, for example, some or all of the exterior surfaces 104, 106. Because the vehicle 100 may be embodied as an autonomous vehicle, a windshield and / or windows may not be required.

[0021] As further shown in FIG. 1, the vehicle 100 is transported by wheels 108 (the vehicle has four wheels, two of which are shown in FIG. 1). Although the exemplary vehicle 100 has four wheels, the systems and methods described herein may be incorporated into vehicles having a fewer or greater number of wheels and / or tracks. Some of the wheels 108 may include protective covers, such as bumpers or fenders (shown in FIG. 2). Such protective covers may be part of the body 102 and / or the exterior surfaces 104, 106. The exemplary vehicle 100 may have four-wheel steering and may generally operate with equal performance characteristics in all directions, e.g., in a first direction in which a first end is at the leading edge of the vehicle 100, and in a second direction opposite the first direction in which a second end is at the leading edge of the vehicle 100. The vehicle 100 may be powered by one or more internal combustion engines, one or more batteries and electric motors, hydrogen fuel cells, or any combination thereof (none of which are shown).

[0022] The exemplary vehicle 100 includes one or more first sensors 110 mounted or otherwise coupled to the body 102 of the vehicle 100. Data generated by the first sensors 110 are used to sense low-energy impact collisions with the vehicle 100, as described in further detail herein. The first sensors 110 associated with the vehicle 100 may include one or more types of modalities. For example, the first sensors 110 may include electromechanical sensors, such as impact sensors (e.g., fiber optic sensors, pressure tubes, piezoelectric or resistive thin film sensors), inertial sensors (e.g., inertial measurement units (IMUs), accelerometers), and / or proximity sensors (e.g., capacitive sensors, microwave sensors, optical / infrared sensors, or ultrasonic proximity sensors). The first sensors 110 may also include resistive, capacitive, fiber optic, piezoelectric or electret microphones, directional microphones, and / or acoustic sensors, such as surface acoustic wave (SAW) devices. The acoustic sensors can be configured as a Wireless Acoustic Sensor Network (WARS) or as a traditional (wired) array.

[0023] The vehicle 100 may also include one or more second sensors 112 mounted on or otherwise coupled to the body 102 of the vehicle 100. In examples of the present disclosure, data generated by the second sensors 112 enables autonomous or semi-autonomous vehicle operation, for example, vehicle navigation, obstacle avoidance, collision handling, etc. The second sensors 112 may include a number of modalities, including, but not limited to, light detection and ranging (LIDAR) sensors, RADAR sensors, ultrasonic transducers, sonar sensors, position sensors (e.g., global navigation satellite system (GNSS) (including global positioning systems (GPS)), compasses, etc.), inertial sensors (e.g., inertial measurement units (IMUs), accelerometers, magnetometers, gyroscopes, etc.), cameras (e.g., red-green-blue (RGB), infrared (IR), intensity, depth, etc.), time-of-flight sensors, microphones, wheel encoders, and / or environmental sensors (e.g., temperature sensors, humidity sensors, light sensors, pressure sensors, etc.). Although the exemplary modality of the first sensor 110 provided above is different from the exemplary modality of the second sensor 112, in some examples, the first sensor 110 and the second sensor 112 may include the same modality and / or the same sensor. This description may distinguish between the first sensor 110 as a sensor that generates data used for a first purpose, e.g., to determine a low-energy impact collision, and the second sensor 112 as a sensor that generates data used for a second (different) purpose, e.g., for normal driving operations. However, in some examples, the first sensor 110 and the second sensor 112 may be the same sensor and / or sensor modality. Without limitation, data from a single sensor may be used for a first purpose as well as a second purpose, such that a sensor may be associated with both the first sensor 110 and the second sensor 112.

[0024] The vehicle 100 may be equipped with a safety system (not shown) that may include airbags, air curtain systems, seat belt tensioners, and the like. For example, in a conventional safety system, the second sensor 112 may include an accelerometer disposed proximate the front of the vehicle 100 and generally along the longitudinal axis X. When the front of the vehicle 100 impacts an object, for example, in a frontal collision, the accelerometer senses a dramatic change in acceleration and an airbag is deployed. In general, conventional safety systems include any vehicle impact mitigation system, which may be an interior safety system designed to directly interact with passengers and / or cargo carried by the vehicle, and / or an exterior safety system, which may be designed to be deployed or otherwise actuated outside the vehicle to protect the vehicle and / or passengers / cargo within the vehicle. Interior safety systems may include, for example, one or more airbags, air curtains, restraint systems, and the like. Exterior safety systems may include exterior bladders, active bumper devices, and the like.

[0025] The conventional safety systems described are generally used to mitigate damage to the vehicle 100, to passengers within the vehicle 100, and / or to cargo within the vehicle 100 in relatively high impact collisions. For example, lower impact or relatively minor collisions, such as fender benders, parking lot accidents, curb bumps, objects striking the body 102 of the vehicle 100, no speed or low speed collisions, etc., may not be detected using these conventional safety systems. Indeed, in some instances, it may not be desirable to use conventional safety systems in response to these relatively minor collisions. For example, it may not be desirable to deploy an airbag in response to a wind-blown shopping cart hitting the door of the vehicle 100.

[0026] Although conventional safety systems are typically used to detect and / or respond to higher impact collisions, it may still be desirable to identify lower impact collisions. For example, these lower impact collisions may still result in damage to the vehicle 100, damage to objects striking the vehicle 100, etc. Aspects of the present disclosure are directed to identifying these low impact collisions.

[0027] In some examples, a collision may be characterized based on the impact energy level of the collision. For example, a collision having an impact energy of the collision equal to or greater than a threshold level (e.g., 5000 Joules) may be a "high impact" or "relatively high impact" collision, while a collision having an impact energy below the threshold level may be a "low impact" or "relatively low impact" type collision. A low energy impact collision may result from a low or medium speed collision with a medium mass object, such as a collision with a bicycle or scooter, a low or medium speed collision with a low mass object, such as an animal, pedestrian or small object, or a low or no speed collision with a high mass object, including a barrier in a parking lot or another slow moving or stationary vehicle. As already explained, a relatively high impact collision may be detected and responded to by a conventional safety system of the vehicle 100. However, a conventional safety system may not detect a low energy impact collision because conventional impact mitigation systems, such as airbags and seat belts, do not need to be used to ensure passenger safety in the event of such a low energy impact collision. However, a low energy impact collision may require the vehicle to respond and perform some other action.

[0028] Aspects of the present disclosure are directed to improved collision detection and response systems in such low-impact collision examples. More specifically, Figure 1 illustrates a scenario in which an object 114 (in an example, a ball) collides with the vehicle 100 as the vehicle 100 traverses an environment. The object slams into the body 102 of the vehicle 100 at a collision location 116. As will be appreciated, the object 114 colliding with the vehicle 100 is merely shown by way of example, and other examples of objects and / or low-impact collisions are provided herein and will otherwise be understood with the benefit of the present disclosure.

[0029] 1, an example of the present disclosure includes a low-energy impact collision detection system 118. The low-energy impact collision detection system 118 is communicatively coupled to a first sensor 110 of the vehicle 100 and one or more system controllers 120. As described in more detail below, the low-energy impact collision detection system 118 may receive sensor data 122 from the first sensor 110 and determine a control command 130 based at least in part on the sensor data 122. The control command 130 indicates an action to be taken, such as, for example, stopping the vehicle, communicating with a teleoperator, storing operational data, etc. The system controller 120 of the vehicle 100 performs the action based on the received control command 130.

[0030] System controller 120 may communicate with and / or control one or more drive systems and / or corresponding systems of other components of autonomous vehicle 100. For example, system controller 120 may control steering, propulsion, braking, safety, emitter, communication, and / or other systems of vehicle 100.

[0031] The low-energy impact collision detection system 118 may generally implement various techniques for detecting and / or responding to collisions of low-energy impacts involving the vehicle 100 and external objects 114 in an environment. As shown in FIG. 1 , the low-energy impact collision detection system 118 may include a collision location estimation component 124, an event detection component 126, and a command generation component 128. For example, the collision location estimation component 124 of the low-energy impact collision detection system 118 may process the sensor data 122 from the first sensor 110 to generate an estimate of the collision location, e.g., the collision location 116, on the body 102 of the vehicle 100. When the first sensor 110 includes different types of sensor modalities, e.g., mechanical and acoustic, the collision location estimation component 124 may process the different types of sensor data 122 from the different modalities of the first sensor 110 separately. For example, if the first sensor 110 includes one or more electromechanical sensors and one or more acoustic sensors, the crash location estimation component 124 may generate a first estimated crash location based on data generated by the one or more electromechanical sensors in response to contact of the object 114 with the vehicle 100 and may generate a second estimated crash location based on data generated by the one or more acoustic sensors in response to contact of the object 114. In some examples, the crash location estimation component 124 may also determine a probability, certainty, or similar metric associated with the estimated location. Examples of determining an estimated location are described in further detail below, including examples related to FIG. 2.

[0032] The sensor data 122 used by the crash location estimation component 124 may be from multiple sensor modalities. Many types of sensor systems inherently exhibit both false positives, which indicate conditions where a sensor signal did not occur, and false negatives, which do not indicate conditions where a sensor signal did occur. Additionally, conditions within the environment in which the vehicle 100 operates may cause the first sensor 110 to generate sensor data that may cause the low energy impact crash detection system 118, e.g., via the event detection component 126, to falsely identify an event. For example, an acoustic sensor may exhibit a large signal due to noise associated with a door being closed, an engine backfiring, a horn honking, the vehicle 100 going through a pothole, etc. Similarly, in an environment with high ambient noise, it may be difficult to isolate sounds associated with an impact and / or generate an accurate estimated location. By deploying different types of sensors and analyzing the data from those sensors, false positives and false negatives may be reduced within the system as disclosed herein.

[0033] For example, the estimated crash location from each type of sensor modality, as determined by crash location estimation component 124, may or may not correspond to the same location on body 102 of vehicle 100. In some examples, sensor data 122 from one or more of sensors 110 of a first modality may indicate a location of a crash, e.g., no crash, while other sensor data 122 of one or more of sensors 110 may indicate a crash. In other examples, due to sensor location, sensor fidelity, sensor accuracy, and / or other factors, the estimated crash location associated with a crash may vary based on different modalities.

[0034] The event detection component 126 of the low-energy impact crash detection system 118 identifies an event based on the estimated crash location(s) determined by the crash location estimation component 124. For example, the event detection component 126 may receive information regarding the estimated crash location and, in some examples, information regarding the certainty of those estimates, and determine whether an event, such as a low-impact crash event, has occurred. In one example, the event detection component 126 may determine an event where multiple estimated crash locations are within a predetermined threshold distance. In another example, the event detection component 126 may determine an event when the certainty of the crash location is equal to or greater than a certainty threshold. The event detection component 126 may implement several different logic flows to identify an event, some examples of which are described below with reference to FIGS. 4 and 5.

[0035] The event detection component 126 may also perform additional functions. For example, the event detection component 126 may include functionality to determine whether other crash detection systems of the vehicle 100 indicate that a crash has occurred, e.g., a relatively high impact crash. For example, the low-energy impact crash detection system 118 is useful for detecting relatively low impact crashes, but may also detect relatively high impact crashes. However, these relatively high impact crashes are more likely to be handled using traditional crash detection and / or crash mitigation systems. Thus, the event detection component 126 may not determine an event (or take further action associated with the event) if the event has already been determined via another system. In the example of a low-energy impact crash event, the event detection component 126 may also include additional information characterizing the low-energy impact crash, such as the vehicle speed at the time of the crash, the force associated with the impact, etc.

[0036] The command generation component 128 includes functionality for generating one or more control commands 130 to respond to an event, e.g., a low-energy impact crash event. As described in detail herein, the control commands 130 may configure the system controller 120 and / or other aspects of the vehicle 100 to perform one or more actions, e.g., to stop the vehicle if currently moving, to collect and transmit or store operational data related to the event (including, but not limited to, sensor data, control data, and any intermediate message data generated), and / or to initiate communication with a remote operator or other person in or near the vehicle. In yet another example, the control commands 130 may configure the vehicle computing system to generate and / or store data with other sensors, e.g., a second sensor 112 that may provide additional information regarding the impact. Without limitation, the control commands 130 may configure a camera or other imaging device with a field of view determined to generate and / or store data that may provide further information regarding the crash event. For example, the control command 130 may cause the vehicle 100 to store data generated by one or more of the second sensors 112 for a predetermined period of time before and / or after the collision, e.g., 30 seconds, 90 seconds, etc.

[0037] In an example where a low-energy impact crash event is detected, the command generation component 128 may issue a control command 130 to bring the vehicle to a stop. The system controller 120 may respond to this control command by decelerating the vehicle 100 to a stop, e.g., a gentle stop associated with a deceleration rate less than a maximum deceleration rate possible for the vehicle 100. For example, a gentle stop may include braking at a particular rate, such as a predetermined rate, a rate less than a first rate (e.g., a maximum braking rate), a rate greater than a second rate (e.g., a minimum braking rate), a rate from among a plurality of available rates that is substantially in the middle (e.g., a third rate from among five rates), a minimum rate from among a plurality of rates, etc. The deceleration rate may refer to a unit of measure over time, such as meters or feet per second squared (m / s^2). In one example, a gentle stop may include decelerating at a rate of 5 feet or 10 feet per second until the autonomous vehicle 100 comes to a stop.

[0038] The command generation component 128 may also respond to the low-energy impact crash event by issuing a control command 130 to initiate communication with a remote operator. The system controller 120 may respond to this control command by enabling a communication channel between the vehicle's communication unit and the remote operator, for example, over a network. The communication channel may be used by the remote operator to communicate with a person(s) located in and around the vehicle 100. This may include passengers traveling in the vehicle 100 at the time of the low-energy impact crash, police officers responding to the scene of the crash, and other people present in the vicinity of the vehicle at the time of the crash.

[0039] Additionally, the command generation component 128 may issue control commands 130 to store or transmit vehicle operation data after a low-energy impact crash event. In response, the system controller 120 may collect and store vehicle operation data for a time interval before and after the event. The vehicle operation data may include vehicle speed, heading, steering angle, as well as sensor data 122 from sensors 110 located on the vehicle 100. The operation data may be stored locally in an on-board memory of the vehicle 100 or transmitted over a network to an external computing system.

[0040] If the event detection component 126 generates a no-impact event, the command generation component 128 may not generate a control command 130 and no action may be taken by the system controller 120. However, in some examples, the command generation component 128 may generate control command(s) to store the sensor data 122 and / or other operational data, for example, if there is a false positive in some of the sensor data 122. In response, the system controller 120 may collect and store the data as described above.

[0041] The vehicle 100 may also include a perception system 132, which in some examples can provide additional (optional) input to the low energy impact collision detection system 118. The perception system 132 generates perception data 134 associated with the environment of the vehicle 100. For example, the perception data 134 can include information about objects, such as the object 114, detected in the environment proximate to the vehicle 100 using sensor data from the second sensor 112. The perception data 134 can include one or more attributes associated with the detected object(s). These attributes can include, but are not limited to, a classification of the object as an object type (e.g., car, pedestrian, cyclist, etc.), a position of the object (e.g., x, y, and / or z position in a local or global coordinate system, relative position, etc.), an orientation of the object, a velocity of the object, an extent of the object (e.g., a size that may be a bounding box representing the object), a trajectory associated with the object, a likelihood of a collision with the object based on the vehicle's position and trajectory, and / or an expected collision location of the vehicle 100 and the object. In some examples, the trajectory of the object can be determined from RADAR processing and / or other techniques.

[0042] The sensory data 134 may be used by the low energy impact crash detection system 118 to improve the reliability of determining an event. For example, the crash location estimation component 124 may generate an estimated crash location based on the sensory data 134 in addition to an estimated location determined based on the sensor data 122. In other examples, the sensory data 134 may indicate that an object is within a threshold distance of the vehicle, e.g., 1 meter, 0.1 meters, etc. The event detection component 126 may use this proximity data to confirm the likelihood of an event, for example, as described below with reference to Figures 4 and 5.

[0043] Additional details of a low energy impact crash detection system useful in a vehicle, including a low impact sensor system disposed on the vehicle 100, will now be described with reference to the remaining figures.

[0044] 2 illustrates an example vehicle 200 having multiple sensors. In an example, vehicle 200 may be similar to or the same as vehicle 100 described above. The sensors may be similar to or the same as first sensor 110 and / or second sensor 112 of FIG. 1. The sensors are configured to generate data associated with a low-energy impact collision involving vehicle 200.

[0045] As shown in FIG. 2, the vehicle 200 includes a body 202, which may be the body 102 described above with reference to the vehicle 100. The body 202 generally defines the surfaces exposed to the exterior of the vehicle, e.g., the environment in which the vehicle 200 is operating. The body 202 generally includes a first exterior surface 204 at a first end and a second exterior surface (not visible in FIG. 2) at a second end opposite the first end. The body 202 also includes sides or lateral sides 206 (one of which is shown) that extend between the first and second ends of the vehicle 200. The vehicle 200 is transported by wheels 208 (two of which are shown in FIG. 2). Some of the wheels 208 may have protective wheel covers, as shown in FIG. 2. Such protective covers may be a portion of the body 202 and / or the exterior surfaces 204, 206. More or fewer body sections may also be included. For example, the body 202 may also include one or more of a windshield, door panels, bumpers, and the like.

[0046] The exemplary vehicle 200 includes one or more sensors mounted or otherwise coupled to the body 202 of the vehicle 200. The sensors may correspond to the first sensor 110 and / or the second sensor 112. Sensor data generated by the sensors may correspond to the sensor data 122 and is used to sense low-energy impact collisions with the vehicle 200, as described in more detail herein. In the example of FIG. 2, the sensors include a first electromechanical sensor 210, a second electromechanical sensor 212, and a third electromechanical sensor 214. The sensors also include a plurality of acoustic sensors including a first acoustic sensor 216a, a second acoustic sensor 216b, a third acoustic sensor 216c, a fourth acoustic sensor 216d, a fifth acoustic sensor 216e, and a sixth acoustic sensor 216f. Two or more of the first acoustic sensor 216a through sixth acoustic sensor 216f and / or other acoustic sensor(s) not shown in FIG. 2 may be referred to herein as “acoustic sensor 216.” The electromechanical sensors 210, 212, 214 and / or acoustic sensor 216 may be arranged in a predetermined pattern to provide a desired coverage area on the body 202 of the vehicle 200. The sensors may be arranged in a pattern that allows sensing of approximately 360 degrees around the vehicle 200. This allows the vehicle 200 to detect a crash at the first end 204 or second end, the lateral side 206, or the wheels 208 of the vehicle 200. A higher density of sensors may be placed in areas of the vehicle that are more likely to be involved in a crash than other areas of the vehicle, such as the lower part of the body 202. In some examples, a portion of the vehicle may not include any sensors. For example and without limitation, the roof may simply be electromechanical and / or acoustic sensors.

[0047] In an example, the electromechanical sensors 210, 212, 214 may be thin film resistive sensors that change conductivity in response to bending or flexion. The electromechanical sensors 210, 212, 214 may be attached to the body 202 of the vehicle 100 to sense strain on the body 202 caused by an impact, such as an impact caused by the object 114. Without limitation, the electromechanical sensors 210, 212, 214 may be fixed to body panels that make up the body 202. For example, the electromechanical sensors 210, 212, 214 may be bonded to an interior surface (opposite the visible, exposed, exterior surface) of a portion of the body 202. Other types of electromechanical sensors may be used in place of or in addition to the resistive sensors shown. For example, additional or other electromechanical sensors may include other shock sensors (e.g., fiber optic sensors, pressure tubes, piezoelectric thin film sensors), capacitive paint, electrodes, inertial sensors (e.g., inertial measurement units (IMUs), accelerometers), and / or proximity sensors (e.g., capacitive, microwave, optical / infrared, or ultrasonic proximity sensors). Electromechanical sensors 210, 212, and 214 may be of the same type or of different types.

[0048] In the illustrated embodiment, the first electromechanical sensor 210 is disposed on a lateral side 206 of the vehicle 200, the second electromechanical sensor 212 is disposed on an end 204 of the vehicle 200, and the third electromechanical sensor 214 is disposed on a wheel 208 cover of the vehicle 200. As illustrated in FIG. 2, the electromechanical sensors 210, 212, 214 may include multiple separate sensors in a predetermined array or arrangement. For example, the enlarged portion 220 of FIG. 2 shows the first electromechanical sensor 210 in more detail. As illustrated, the first electromechanical sensor 210 includes multiple sensor strips including a first sensor strip 210a and a second sensor strip 210b that are generally horizontally arranged and vertically spaced apart. The first electromechanical sensor 210 also includes a third sensor strip 210c, a fourth sensor strip 210d, a fifth sensor strip 210e, and a sixth sensor strip 210f that are generally vertically arranged and horizontally spaced apart. The strips 210a-210f may be arranged to provide coverage of the lateral side 206 of the vehicle 200. The illustrated arrangement is merely for example and other patterns are contemplated. In the example of FIG. 2, the electromechanical sensors 212 are also shown as horizontal and vertical strips arranged to cover the first end 204 of the vehicle. The sensors are located in areas of the vehicle that are more likely to be involved in a crash, such as the lower part of the body 202. Although individual segments arranged in strips are shown, the lateral side 206 and / or end 202 may also be covered by the sensor array.

[0049] The arrangement of the first electromechanical sensor 210 and the second electromechanical sensor 212 is merely an example. Other patterns of the strips 210a-210f are contemplated, which may include, for example, more or fewer strips. Additionally, while the strips 210a-210f are generally arranged horizontally and vertically, other examples may include angles, overlaps, and / or other configurations. Each of the strips 210-210f may include, for example, multiple separate sensors spaced along the length of the strip and / or the width of the strip, or may be a single sensor. As detailed herein, by including several separate sensors, a more accurate estimate of the impact location with the vehicle 200 may be determined.

[0050] 2, the vehicle 200 also includes acoustic sensors 216. The acoustic sensors 216 may be microphones, such as, for example, resistive, capacitive, fiber optic, piezoelectric or electret microphones, and / or SAW devices. The microphones may have a sensitivity, frequency bandwidth, and / or dynamic range suitable for capturing sounds generated by a typical collision with the vehicle 200 by any object. Each of the acoustic sensors 216 may be individual sensors of the same or different types, or the acoustic sensors 216 may be configured as a Wireless Acoustic Sensor Network (WARS) or as a conventional (wired) array.

[0051] As shown, the acoustic sensors 216 are distributed around the vehicle 200. For example, the first acoustic sensor 216a through the fourth acoustic sensor 216d are spaced apart from one another and coupled to the lateral side 206. The fifth acoustic sensor 216e and the sixth acoustic sensor 216f are shown disposed at the first end 202 of the vehicle 200. In the illustrated example, the acoustic sensors 216 are illustrated as being spaced apart only in a generally horizontal direction, but in other examples, the spacing can be in other and / or additional directions. The acoustic sensors 216 can also include more or fewer sensors. In an example, the acoustic sensors 216 include multiple spaced apart sensors disposed on the body 202 of the vehicle 200, for example, on each side. In an example, the acoustic sensors 216 generate signals in response to environmental noise, and signals from the multiple acoustic sensors 216 can be used to localize the location of the sound source, as described herein. In some examples, the acoustic sensors 216 may be numbered and / or positioned to detect sounds originating on any exterior surface of the vehicle 200. Additionally, the acoustic sensors may be positioned to at least partially overlap the areas of the body 202 monitored by the electro-mechanical sensors 210, 212, 214.

[0052] Portions of vehicle 200 not shown in Figure 2 may have sensors in the same distribution as the sides 206 and ends 204 shown, i.e., the location of the vehicle and sensors may be substantially symmetrical about the lateral and longitudinal axes, as in bidirectional vehicle 100. In other embodiments, the first or leading end 204 of vehicle 200 may differ and / or include a different distribution of sensors than the second or trailing end, as in conventional vehicles. Although Figure 2 shows the placement of sensors on vehicle 200, other types and placements of sensors are contemplated.

[0053] 2 include electromechanical and acoustic sensors, but other types of sensors capable of detecting impacts or side effects resulting from impacts can also be incorporated into the same framework. For example, ultrasonic and optical / infrared sensors for detecting the proximity of an object, as well as fusion sensors including both an accelerometer and a microphone, can also be used in place of or in addition to the mechanical and acoustic sensors described above.

[0054] Each of the electromechanical and / or acoustic sensors may have a unique identifier. The low energy impact crash detection system 118 may associate a specific location with each sensor indicating the location of the sensor with respect to the vehicle or the section of the vehicle on which it is located. The location associated with each sensor may be represented in various ways in two or three dimensions, such as, by way of non-limiting example, the three dimensional coordinates of the center of the sensor relative to the center of the vehicle 200, the two dimensional coordinates of the center of the sensor relative to an origin on a plane representing a cross section of the vehicle 200 (e.g., the lateral side 206), or a set of coordinates indicating the extent of the sensor in either two or three dimensions. The low energy impact crash detection system 118 may also use an internal representation that captures the spatial relationship of the sensors in two or three dimensions, such as a grid representation with grid cells indicating the presence / absence of a sensor or sensor segment at a grid location, and / or a signal value from a sensor located within a grid cell.

[0055] For example, each electromechanical sensor segment 210a-210f is uniquely identified and associated with a particular location on the body 202 of the vehicle 200, such as 210a on the lateral side 206 and associated with the topmost location closest to the first end 204. The sensor data 122 received by the low energy impact crash detection system 118 includes a sensor identifier and its corresponding signal value for each of the electromechanical sensors 210, 212, 214 and acoustic sensor 216 located on the vehicle 200.

[0056] As an example, FIG. 2 illustrates an impact location 218 that may be the impact location 116. Although the impact location 218 is illustrated on the lateral side 206, it is understood that the impact location 218 may be on any side of the vehicle or on the wheels 208 of the vehicle 200. As illustrated, there may not be a sensor(s) located at the exact impact location 218, or the impact location 218 may correspond to one or more sensor locations. As a result of the impact at the impact location 218, the sensor data 122 from the various electromechanical sensors 210, 212, 214 and / or the various acoustic sensors 216 may exhibit elevated signal values. For example, the electromechanical sensor 210 may generate a higher signal value than the sensor 212 located further away from the impact area. Additionally, each of the strips including the electromechanical sensors 210 may generate a different signal based on their proximity to the location. Similarly, the second and third acoustic sensors 216b, 216c, which are the acoustic sensors 216 closest to the crash location 218, are expected to generate higher signal values ​​than, for example, the fifth and sixth acoustic sensors 216e, 216f, which are located further away from the crash location 218. The crash location estimation component 124 of the low energy impact crash detection system 118 estimates the crash location 218 based on the sensor signal values ​​from the sensors 210, 212, 214, 216 included in the sensor data 122. For example, as described in more detail below, the crash location estimation component 124 can determine a first estimated crash location 220 based on sensor data from the electromechanical sensors 210, 212, 214 and / or can estimate a second estimated location 222 based on sensor data from the acoustic sensor 216.

[0057] In the example shown in FIG. 2, the electromechanical sensors 210, 212, 214 cover a substantial portion of the exterior surfaces 204, 206, 214 of the vehicle 200. In the illustrated example, no sensor is shown exactly at the impact location 218. However, in other examples, the impact can be located directly at the sensor location. In general, the value of the sensor signal generated by the electromechanical sensors 210, 212, 214 varies depending on the proximity of the impact location to the sensor location. For example, in the impact example, the sensor (if any) at the impact location 218 records the highest signal value, and the signal value decreases as the distance from the impact location 218 increases. In the example shown in the enlarged portion of FIG. 2, the signal value is highest in the first and fifth strips 210a, 210e and lowest in the third strip 210c. The sensor signal value in non-impact areas (e.g., not involved in the impact) of other sides of the vehicle 200, e.g., the sensor signal value at sensor 212, may be much lower or even zero. The signal value may indicate the energy of the impact, the degree of distortion of the flexible sensor, or any other indication of the severity of the impact. The sensor characteristics of the deployed sensor may provide a mapping between the signal level and the severity of the impact. In some examples, the signal level may be used to classify the impact as a relatively high impact impact or a low energy impact impact.

[0058] The crash location estimation component 124 of the low energy impact crash detection system 118 may implement various methods to estimate the location of the crash, e.g., the first estimated crash location 220 of FIG. 2, from the electromechanical sensors 212, 214, 216. In one non-limiting example, the sensor signal values ​​may be clustered into two groups, e.g., by using k-means with k=2. In this example, the cluster with the higher mean value corresponds to sensors in the crash region and the cluster with the lower mean value corresponds to sensors in the non-crash region. The first estimated crash location 220 may be estimated as the spatial center of the sensor cluster with the higher mean value. This method avoids the problem of determining a threshold setting to distinguish signals indicative of a crash from signals indicative of a non-crash region. However, in other examples, a predetermined threshold value may be used.

[0059] For example, sensor signal levels from non-crash regions may be learned from training data collected during typical vehicle operation. The learned signal levels corresponding to the non-crash regions may be used as reference signals. For example, the measured sensor signal may be compared to the reference signal, and if the difference is higher than a threshold, the sensor may be identified as being in the crash region. The estimated crash location may be calculated as the spatial center of the sensor identified as being in the crash region. In another example, the estimated crash location may correspond to the location of the sensor with the highest signal level.

[0060] In another example, the crash localization component 124 may use an internal grid representation to represent the spatial distribution of the sensors. Although the electromechanical sensors 210, 212, 214 are shown as strips, they may also be represented as individual sensors associated with separate regions or cells on the grid. For example, a portion of the surface of the vehicle 200, such as the lateral side 206, may be represented as a two-dimensional grid. In an example, a unique identifier for the sensor 210 may be associated with the sensor's location on the grid. The identifier may encapsulate location information, as a non-limiting example. In another example, a lookup table may be available to map the unique identifier to a location on the grid representing the lateral side 206 of the vehicle 200, such as the sensor 210a in row x and column y of the grid. The grid cells may represent individual sensors, or a single sensor may occupy multiple grid cells. The size of the grid cells may vary based on the granularity of the crash localization of the low-energy impact crash detection system 118. All grid cells do not have to be the same size; for example, the grid may be finer (having smaller cells) in areas with denser sensor coverage or in areas where finer positioning of collisions is desired.

[0061] The crash location estimation component 124 may also implement techniques to determine a second estimated crash location 222 from the sensor data generated by the acoustic sensors 216. The acoustic portion of the sensor data 122 includes the signal levels captured by each acoustic sensor 216 and the corresponding unique identifier of the sensor. The unique identifier is associated with the location of the sensor on the body 202 of the vehicle 200. The crash location estimation component 124 can use several techniques to locate the source of the sound or calculate an estimate of the source location, for example, as the second estimated crash location 222, from the sound waves measured at different points. In this case, the source of the sound is at the crash location 218, and locating the source of the sound provides the second estimated crash location 222. Localization techniques include source location estimates based on energy levels or received signal strength (RSS), time of arrival (TOA) measurements, time difference of arrival (TDOA) measurements, and direction of arrival measurements at multiple spatially distributed acoustic sensors. Signal processing techniques such as steered response power (SRP) functions may also be used for source localization using input from spatially distributed acoustic sensors. In one example, using data from acoustic sensors 216b and 216c, an estimate of the crash location may be determined, for example, at second estimated location 222 shown in FIG. 2. In an example, crash location estimation component 124 may identify data associated with sounds that exceed a threshold level, e.g., sounds that exceed a threshold decibel level, for the crash location estimation. In another example, crash location estimation component 124 may identify data associated with sounds that vary beyond a threshold from an ambient or baseline level of sound, such that, for example, the actual decibel level varies based on ambient noise.

[0062] In the example of FIG. 2, the first estimated impact location 220 may be relatively close to the actual impact location 218, and the second estimated impact location 222 may be relatively far from the actual impact location 218. For example, the first estimated impact location 220 may have higher fidelity because there are more electromechanical sensors in a finer array. Alternatively, the acoustic sensors may be arranged in a line, and the second estimated impact location may be generally along that line. Of course, different arrangements may provide different results. In addition to the arrangement and number of sensors, the accuracy of the first estimated impact location 220 and / or the second estimated impact location 222 may vary based on the quality of the sensor data, the processing techniques used, and / or other factors. 2 shows the first and second crash locations 220, 222 as being considered separately using different sensor data types, in other examples, the crash location estimation component 214 may determine an estimated crash location based on, for example, sensor data from the electromechanical sensors 210, 212, 214 and sensor data from the acoustic sensor 216. Without limitation, a model may be trained on the acoustic data and data from the electromechanical sensors to determine a single estimated crash location. In further examples, the single estimated crash location may be an average or other function of the first estimated crash location 220 and the second estimated crash location.

[0063] In some examples, the collision location estimation component 124 may also generate a third estimated collision location 224. The third estimated collision location 224 may be based on sensory data, such as, but not limited to, the sensory data 134. As shown in FIG. 2, the third estimated location may be an area where there is a higher probability of collision based on the object(s) and / or corresponding trajectories of the objects shown in the sensory data 134. For example, the third estimated collision location 224 may correspond to an area where an object proximate to the vehicle 200 and / or an object based on the trajectory of the object may collide with the vehicle 200. In some examples, the sensory data 134 may include an expected collision area corresponding to the third estimated location 224.

[0064] The crash location estimation component 124 may also implement various artificial intelligence (AI) techniques to predict the location and type of crash from a set of sensor signals. The AI ​​techniques may include machine learning (also commonly referred to as machine learning models), such as one or more neural networks. For example, a neural network may be trained to classify an input vector including sensor signal values ​​of the sensors 110 into types of crashes (including non-crashes). A separate neural network, such as a convolutional neural network, that encodes the spatial distribution of the input signals may be trained to estimate the crash location from the sensor data 122. A training data set corresponding to low-energy impact crashes may be collected from operational data of a vehicle with associated sensors. The training data set may be labeled with the location and type of the crash and may include data where there is no crash. When the trained model(s) are deployed as an implementation of the crash location estimation component 124, the sensor data 122 forms an input to the trained model, and an estimate of the location and type of the crash is obtained as an output. In an example, separate neural networks each trained with a different type of sensor modality may be deployed, for example, one or more neural networks trained with acoustic sensor data and a set of separate neural networks trained with electromechanical sensor data. Alternatively, a single neural network with aggregated different types of sensor signals as input may also be used. Further examples of architectures include neural networks such as ResNet50, ResNet101, VGG, DenseNet, PointNet, etc. The techniques used to estimate the crash location may be similar or different for each type of sensor.

[0065] Although described in the context of neural networks, any type of machine learning can be used consistent with the present disclosure. For example, machine learning algorithms can include, but are not limited to, 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 net, least-angle regression (LARS)), decision tree algorithms (e.g., classification and regression trees (CART), iterative dichotomiser 3 (ID3), chi-squared automated interaction detection (CHAID), decision stump, conditional decision trees), Bayesian algorithms (e.g., naive Bayes, Gaussian naive Bayes, multinomial naive Bayes, average one-dependence (AODE), etc.), and / or ensemble algorithms (e.g., ensemble algorithms ... estimators), Bayesian belief networks (BNNs), Bayesian networks), clustering algorithms (e.g., k-means, k-medians, expectation maximization (EM), hierarchical clustering), association rule learning algorithms (e.g., perceptrons, backpropagation, Hopfield networks, RBFNs (Radial Basis FunctionNetwork), deep learning algorithms (e.g., Deep Boltzmann Machines (DBM), Deep Belief 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), Sammon Mapping, Multidimensional Scaling (MDS), Projection Pursuit, 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, Stacked Generalization (Blending), Gradient Boosting Machines (GBM), Gradient Boosted Regression Trees (GBRT), Random Forests), SVM (Support Vector Machines), supervised learning, unsupervised learning, semi-supervised learning, etc.

[0066] As mentioned above, the location associated with each sensor may be expressed in various ways in two or three dimensions, such as three-dimensional coordinates of the center of the sensor relative to an origin at the center of the vehicle 200, two-dimensional coordinates of the center of the sensor relative to an origin on a plane representing a section of the vehicle 200 (e.g., side 206), or cell locations on a two or three dimensional grid. The estimated impact locations 220, 222, 224 may be expressed in the same coordinate system using the same or different granularity, or may be expressed using a different scheme. For example, the estimated impact locations 220, 222, 224 may be expressed in terms of an area on the body 202 of the vehicle 200, e.g., first side, second side, first end, second end, third wheel cover, etc. In another example, the estimated impact locations 220, 222, 224 may be expressed at a lower granularity in the same two or three dimensional grid used to describe the sensor locations. The estimated crash locations 220, 222, 224 may also be represented using different schemes based on the level of certainty of the estimate. For example, the third estimated crash location 224 may be represented in terms of an area on the body 202 of the vehicle 200 (e.g., a first side), the second estimated location 222 may be represented as a location on a two-dimensional grid, and the first estimated location 220 may be represented as a location on a two-dimensional grid with a coarser granularity.

[0067] As described, the crash location estimation component 124 may determine a first estimated crash location 220 based on the electromechanical sensor data and may determine a second estimated crash location 222 based on the acoustic data. Additionally, a third estimated crash location 224 may be determined based on the sensory data 134. Although a single estimate of crash location is shown for each sensor modality / data source type, there may be zero, one or multiple crash location estimates from each sensor type.

[0068] The event detection component 126 can use some or all of the first, second, and third estimated locations 220, 222, 224 to determine if a collision event has occurred. In some examples, the event detection component 126 compares the first estimated collision location 220 and the second estimated collision location 222. The comparison can be in the form of a spatial distance between the two locations 220, 222. If the distance is below a threshold, the two locations are determined to correspond to one another. If the distance between the collision locations 220, 222 is equal to or greater than a threshold, the locations 220, 222 are determined to not correspond to one another and a no-impact event is determined by the event detection component 126. For example, if the third estimated location 224 is available based on the sensory data 134, a collision event may be determined by the event detection component 126 when the first estimated collision location 220 and / or the second estimated collision location 222 overlap the area covered by the third estimated collision location 224. Otherwise, a non-crash event is determined by the event detection component 126. If the sensory data 134 is not available, the event detection component 126 determines a collision event where the two locations 220 and 222 are determined to correspond to one another. If either the electromechanical sensor or the acoustic sensor does not generate an estimate of the location of the collision, for example, if a collision is not detected by the sensor, the event detection component 126 determines a no-crash event. In examples, there may not be a threshold used to determine correspondence, instead, correspondence between two positions 220, 222 may be based on both positions being on the same surface of the body 202 of the vehicle 200, e.g., a first side, a second side, a first end, a second end, a third wheel cover, etc.

[0069] In another embodiment of the event detection component 126, a third estimated crash location 224 based on the sensory data 134 is received as an input from the crash location estimation component 124. As described above, an estimated location(s) of the crash 220, 222 is also obtained for each type of sensor modality, e.g., electromechanical, acoustic. The estimated crash location 224 is compared to the estimated crash locations 220, 222. The comparison may be in the form of determining an overlap between the location 224 and another location, e.g., 220 or 222. If there is an overlap between the location 224 and any one of the locations 220 or 222, the event detection component 126 determines a crash event. Otherwise, if there is no overlap, e.g., both the locations 220 and 222 are outside the locations shown in 224, a non-crash event is determined by the event detection component 126. As will be described, using a combination of estimated locations from different sensor modalities and sensory data allows a crash event to be identified with higher certainty.

[0070] FIG. 3 illustrates a block diagram of an example system 300 for implementing the techniques described herein. In some examples, the system 300 may include a vehicle 302 that may correspond to the vehicle 100 of FIG. 1 and / or the vehicle 200 of FIG. 2. In some examples, the vehicle 302 may be an autonomous vehicle configured to operate according to a Level 5 classification issued by the National Highway Traffic Safety Administration, which describes a vehicle capable of performing all safety-critical functions for the entire journey without expecting a driver (or passenger) to constantly control the vehicle. However, in other examples, the autonomous vehicle 302 may be a fully or partially autonomous vehicle having other levels or classifications. Additionally, in some cases, the techniques described herein may be usable by non-autonomous vehicles.

[0071] Vehicle 302 may include a vehicle computing device 304, one or more sensor systems 306, one or more emitters 308, one or more communication connections 310, at least one direct connection 312 (e.g., for physically coupling to vehicle 302 to exchange data and / or provide power), and one or more drive systems 314.

[0072] In some examples, the sensor system 306 may include LIDAR (light detection and ranging) sensors, RADAR sensors, ultrasonic transducers, sonar sensors, position sensors (e.g., GPS, compass, etc.), inertial sensors (e.g., inertial measurement units (IMUs), accelerometers, magnetometers, gyroscopes, etc.), cameras (e.g., RGB, IR, intensity, depth, time of flight (TOF), etc.), microphones, wheel encoders, environmental sensors (e.g., temperature sensors, humidity sensors, light sensors, pressure sensors, etc.), and the like. The sensor system 306 may include multiple instances of each of these or other types of sensors. For example, the LIDAR sensors may include individual LIDAR sensors located at corners, front, rear, sides, and / or top of the vehicle 302. As another example, the cameras may include multiple cameras positioned at various locations on the exterior and / or interior of the vehicle 302. The sensors 306 may provide input to the vehicle computing device 304.

[0073] The vehicle 302 may also include emitters 308 for emitting light and / or sound, as described above. The emitters 308 in this example may include interior audio and visual emitters for communicating with passengers of the vehicle 302. By way of example and not limitation, the interior emitters may include speakers, lights, signs, display screens, touch screens, haptic emitters (e.g., vibration and / or force feedback), mechanical actuators (e.g., seat belt tensioners, seat positioners, head rest positioners, etc.), and the like. The emitters 308 in this example may also include exterior emitters. By way of example and not limitation, the exterior emitters in this example may include lighting to signal a direction of travel or other indicators of the vehicle's behavior (e.g., indicator lights, signs, lighting arrays, etc.), as well as one or more audio emitters (e.g., speakers, speaker arrays, horns, etc.) to acoustically communicate with pedestrians or one or more other nearby vehicles, including acoustic beam steering technology.

[0074] Vehicle 302 may also include communications connections 310 that enable communication between vehicle 302 and one or more other local or remote computing devices. For example, communications connections 310 may facilitate communication with other local computing devices on vehicle 302 and / or drive system 314. Communications connections 310 may additionally or alternatively enable vehicle 302 to communicate with other nearby computing devices (e.g., other nearby vehicles, traffic signals, etc.). Communications connections 310 may additionally or alternatively enable vehicle 302 to communicate with computing device 316.

[0075] The vehicle computing device 304 may include one or more processors 318 and a memory 320 communicatively coupled to the one or more processors 318. In the depicted example, the memory 320 of the vehicle computing device 304 stores a perception component 322, one or more system controllers 324, and a collision detection system 326 including a high-energy collision detection system 328 and a low-energy impact collision detection system 330. While shown in FIG. 3 as residing in the memory 320 for illustrative purposes, it is contemplated that the perception component 322, the one or more system controllers 324, and the collision detection system 326 may additionally or alternatively be accessible to the vehicle 302 (e.g., stored remotely).

[0076] In some cases, the perception component 322 may include functionality for performing object detection, segmentation, and / or classification. In one example, the perception component 322 may provide processed sensor data indicative of the presence of an entity proximate to the vehicle 302 and / or the classification of the entity as an entity type (e.g., automobile, pedestrian, cyclist, animal, building, tree, road surface, curb, sidewalk, unknown, etc.). In additional or alternative examples, the perception component 322 may provide processed sensor data indicative of one or more characteristics associated with the detected entity (e.g., tracked object) and / or the environment in which the entity is located. In some examples, the characteristics associated with the entity may include, but are not limited to, x-position (global and / or local position), y-position (global and / or local position), z-position (global and / or local position), orientation (e.g., roll, pitch, yaw), entity type (e.g., classification), velocity of the entity, acceleration of the entity, range (size) of the entity, etc. Characteristics associated with an environment may include, but are not limited to, the presence of other entities in the environment, the state of other entities in the environment, time of day, day of the week, season, weather conditions, darkness / light indications, and the like.

[0077] In at least one example, the vehicle computing device 304 may include one or more system controllers 324, which may be configured to control steering, propulsion, braking, safety, emitter, communication, and other systems of the vehicle 302. The system controllers 324 may communicate with and / or control systems corresponding to the drive system 314 and / or other components of the vehicle 302.

[0078] The system controller 324 may be communicatively coupled to one or more sensors of the vehicle's sensor system 306. As a non-limiting example, the sensors may detect the presence of objects in the vehicle's environment and / or determine attributes of those objects. The system controller 324 may also trigger the activation of safety systems of the vehicle 302 when it is determined that the safety systems should be activated. For example, the system controller 324 may instruct an airbag control unit to deploy one or more airbags or send a signal to a tensioner arranged to adjust the tension of one or more restraints. Other safety systems are known and may be enabled. In other embodiments, the system controller 324 may direct the activation of multiple safety systems. In some embodiments, some or all of the functions of the safety system controller 324 may be performed away from the vehicle 302, for example, on a remote server associated with the dispatch or headquarters of the vehicle 302, or in the cloud. In other embodiments, some or all of the functions of the system controller 324 may be performed on the vehicle 302 to minimize any delays that may result from the transmission of data between locations.

[0079] The drive system 314 may include many vehicle systems including a high voltage battery, a motor to propel the vehicle, an inverter to convert direct current from the battery to alternating current for use by other vehicle systems, a steering system including a steering motor and a steering rack (which may be electric), a braking system including hydraulic or electric actuators, a suspension system including hydraulic and / or pneumatic components, a stability control system for braking force distribution to mitigate loss of traction and maintain control, an HVAC system, lighting (e.g., lighting such as headlights / taillights that illuminate the perimeter of the vehicle), and one or more other systems (e.g., cooling systems, safety systems, on-board charging systems, other electrical components such as DC / DC converters, high voltage junctions, high voltage cables, charging systems, charging ports, etc.). Additionally, the drive system 314 may include a drive system controller to receive and pre-process data from sensors and to control the operation of various vehicle systems. In some examples, the drive system controller may include one or more processors and a memory communicatively coupled to the one or more processors. The memory may store one or more modules that perform various functions of the drive system 314. Additionally, drive system 314 also includes one or more communication connections that enable each drive system to communicate with one or more other local or remote computing devices.

[0080] Crash detection system 326 includes a high-energy crash detection system 328 configured to detect relatively high impact crashes using input from one or more sensors in sensor system 306. For example, input from an accelerometer located proximate the front of the vehicle may indicate a frontal crash with an object. A crash indication from high-energy crash detection system 328 causes system controller 324 to deploy vehicle safety systems, such as, for example, one or more airbags, air curtains, restraint systems, etc.

[0081] 1, and its components 332 and 334 may correspond to the crash location estimation component 124 and the event detection component 126, respectively. The low-energy impact crash detection system 330 is configured to detect low-energy impact crashes and is operably coupled to the system controller 324.

[0082] In response to a crash event indication from the low-energy impact crash detection system 330, the system controller 324 may take one or more of the following actions, by way of non-limiting example: The system controller 324 may engage the drive system 314 to stop the vehicle 302. The system controller 324 may open a communication connection 310 between the communication unit of the vehicle 302 and a remote operator via a network 336. This connection may allow two-way communication between the remote operator and a person(s) located in and around the vehicle 302. The system controller 324 may also store vehicle operation data such as vehicle speed, heading, steering angle, as well as data from the sensor system 306 for time intervals before and after the low-energy impact crash event. The operation data may be stored locally in the on-board memory 320 of the vehicle 302 or may be transmitted to the external computing device 316 via the network 336 and stored in an external memory 342. The system controller 324 may also engage emitters 308 on the vehicle 302 to indicate a hazard condition when a low-energy impact crash indication is received. For example, an audio emitter may sound a horn or make an announcement inside the vehicle 302. Visual emitters such as hazard lights on the exterior of the vehicle 302 may also be activated. Lights inside the vehicle 302 may also indicate a crash event.

[0083] In some examples, the vehicle 302 may transmit operational data, including raw data or processed sensor data from the sensor system 306, to one or more computing devices 316 over the network 336. The operational data may be transmitted in response to a crash event detected by the crash detection system 326. In other examples, the vehicle 302 may transmit processed operational data and / or a representation of the operational data to the computing device 316 in near real-time after a predetermined period of time, at a particular frequency. In some cases, the vehicle 302 may transmit the raw data or processed operational data to the computing device 316 as one or more log files.

[0084] In some examples, some or all aspects of the components described herein may include any model, algorithm, and / or machine learning algorithm. For example, the components in memory 320, 342 may be implemented as neural networks. As can be understood in the context of the present disclosure, neural networks may utilize machine learning, which may refer to a broad class of such algorithms in which output is generated based on learned parameters.

[0085] 4 and 5 illustrate exemplary processes according to embodiments of the present disclosure. These processes are illustrated as logical flow graphs, each operation of which represents a sequence of operations that can be implemented in hardware, software, or a combination thereof. In the context of software, the operations represent computer-executable instructions stored on one or more computer-readable recording media that, when executed on one or more processors, perform the recited operations. Generally, computer-executable instructions include routines, programs, objects, components, data structures, etc. that perform particular functions or implement particular abstract data types. The order in which the operations are described is not intended to be construed as limiting, and any number of the described operations can be combined in any order and / or in parallel to implement a process.

[0086] It should be understood that the subject matter presented herein can be implemented as an article of manufacture, such as a computer process, a computer controller, a computing system, or a computer-readable storage medium. Although the subject matter described with respect to process 400 and process 500 is presented in the general context of operations that may be performed on and / or in conjunction with one or more computing devices, those skilled in the art will recognize that other implementations may be performed in combination with various program / control modules. Generally, such modules include routines, programs, components, data structures, and other types of structures that perform particular tasks or implement particular abstract data types.

[0087] Those skilled in the art will also understand that aspects of the subject matter described with respect to process 400 and process 500 may be implemented or combined with other computer system configurations beyond those described herein, including multiprocessor systems, microprocessor-based or programmable consumer electronics, minicomputers, mainframe computers, handheld computers, mobile phone devices, tablet computing devices, dedicated hardware devices, network equipment, and the like.

[0088] 4 is a flow diagram illustrating an example process 400 for controlling a vehicle, such as one of the vehicles 100, 200, 302, based on ascertaining an event. In some examples, the process 400 may be performed by a vehicle controlled by a low energy impact collision detection system 118, 330 and a system controller 120, 324 of the present disclosure.

[0089] More specifically, the example process 400 includes, at operation 402, receiving sensor data from electromechanical sensor(s) disposed on a body of a vehicle. For example, the sensor data may be received from one or more of the electromechanical sensors 110, 210, 212, 214, 306 described herein. The electromechanical sensor may be associated with an exterior surface of the vehicle, for example, a body panel of the vehicle. The electromechanical sensor generates a signal in response to an impact with the vehicle. In some implementations, the sensor data may be a signal value detected by an impact sensor.

[0090] At operation 404, process 400 also includes determining a first estimated crash location. In an example, operation 404 may include comparing signals corresponding to the sensed crash from a plurality of electromechanical sensors and estimating a location based on the signals. Operation 404 may be implemented by the crash location estimation component 124, 332 according to the functionality described above. In some examples, the first estimated location may be the second estimated location 222 of FIG. 2.

[0091] Process 400 may also include receiving sensor data from one or more acoustic sensors at operation 406. For example, the sensor data may be received from acoustic sensor 216 disposed on vehicle 200. In an example, the acoustic sensor may include multiple sensors disposed around a perimeter of the vehicle. The acoustic sensor is configured to generate an audio signal in response to detecting a sound, e.g., a sound associated with a vehicle crash. For example, the intensity or value of the audio signal may correspond to the severity of the impact to the vehicle.

[0092] At operation 408, process 400 includes determining a second estimated crash location. In an example, operation 408 may include comparing the audio signal received in operation 406 to signals of a plurality of acoustic sensors and estimating the second location based on the signals. Operation 408 may be implemented by the crash location estimation component 124, 332 according to the functionality described above. In some examples, the second estimated location may be the first estimated location 220 of FIG. 2.

[0093] The process also includes determining whether the first estimated crash location corresponds to the second estimated crash location in operation 410. For example, operation 410 may compare the first estimated location and the second estimated location to determine whether the locations are within a threshold distance of each other. Operation 410 may be implemented by an event detection component 126, 334 of a low-energy impact crash detection system 118, 330 of the present disclosure.

[0094] If, in operation 410, it is determined that the first estimated crash location corresponds to the second estimated crash location, then, in operation 412, process 400 determines a crash event. For example, process 400 determines a crash when both the first sensor data and the second sensor data confirm an approaching crash. In at least some examples, but not limited to, certain controls may be applied based at least in part on the determination of a crash, such as storing sensor data associated with a set amount of time before and / or after detection, reducing speed, sending one or more information or confirmation requests to a remote operator system, etc.

[0095] Alternatively, if in operation 410 it is determined that the first estimated impact location does not correspond to the second estimated impact location of the impact, a no-impact event is determined in operation 414 of process 400. Specifically, if the sensors determine a collision at a location that is not associated with one another, or if only one of the sensor modalities generates data associated with the collision, process 400 determines that a collision with the vehicle has not occurred. For example, if no event is determined, no action associated with the collision is performed.

[0096] At optional operation 416, process 400 may optionally receive sensory data. For example, the sensory data may be sensory data 134 from perception system 132. As described herein, the vehicle's sensory system may generate data regarding objects in the vehicle's environment using, for example, sensor data from one or more sensors disposed on the vehicle. In examples, the sensory data may be determined from data other than the sensor data received at operations 402, 406 described above.

[0097] In an example where the process 400 includes receiving sensory data, in operation 418, the process 400 includes determining whether the sensory data confirms a crash event. For example, operation 418 may include determining whether the sensory data indicates an object proximate to the vehicle near the first estimated location and the second estimated location. For example, operation 418 may determine that the sensory data confirms a crash event if the object in the sensory data is within a threshold distance (e.g., one meter) of the vehicle and / or within a threshold distance of an estimated location on the vehicle of the crash event determined in operation 412 based on the first estimated crash location and / or the second estimated crash location. Operation 418 may be implemented by an event detection component 126, 334 of a low energy impact crash detection system 118, 330 of the present disclosure.

[0098] If, at operation 418, it is determined that the sensory data confirms a crash event, process 400 includes controlling the vehicle based on the crash event. As detailed herein, operation 420 may include controlling the vehicle to stop, controlling the vehicle to request assistance from passengers, authorities, and / or nearby persons, controlling the vehicle to generate an incident report, controlling the vehicle to request assistance from a remote operator, controlling the vehicle with reduced functionality, such as, for example, a lower maximum speed, by moving forward only in a second direction based on data from fewer sensors, etc. As noted above, operations 416, 418 may be optional, and if the process does not include those operations, process 400 may proceed directly from operation 412 to operation 420, for example, to control the vehicle based on the determined crash event.

[0099] In some examples, operations 416, 418 may serve as further checks or confirmations of the crash detected at the first estimated location and / or the crash detected at the second estimated location. Data other than sensory data may also or alternatively be received and / or considered in conjunction with process 400. For example, the process may also or alternatively include receiving an input from a primary crash detection system indicating whether a relatively high impact crash has occurred. For example, if the vehicle is involved in a more serious accident, the vehicle may be controlled according to a protocol for a larger impact crash. Aspects of the present disclosure may relate to controlling the vehicle in operation 420, for example, in response to a low impact crash, while the functionality may be taken over by another processing system in the case of a higher impact crash. Thus, in some examples, in the absence of an indication of a relatively high impact crash, a crash event corresponding to a low energy impact crash event may be determined in operation 412.

[0100] 4, if in operation 418 it is determined that the sensory data does not confirm a crash event, process 400 may proceed to operation 414 where no crash is determined, as described above. That is, even if in operation 410 it is determined that the first location corresponds to the second location, the optional sensory data may nullify the crash event. In other examples, other actions may be performed in response to a discrepancy in the sensory data of the first location and the second location. For example, the area of ​​the crash location may be flagged and monitored more closely, e.g., using sensors other than the electromechanical sensor(s) and / or acoustic sensor(s). Additionally, additional data regarding the crash location may be requested from sensors other than the electromechanical sensor(s) and / or acoustic sensors.

[0101] 5 is a flow diagram illustrating an example process 500 for controlling a vehicle 100, 200, 302 based on identifying an event. In some examples, the process 500 may be performed by a low-energy impact collision detection system 118, 330 of the present disclosure, and the vehicle may be controlled by a system controller 120, 324 described above.

[0102] More specifically, at operation 502, process 500 includes receiving sensory data. For example, the sensory data may be sensory data 134 from the perception system 132 of the vehicle 100. As described herein, the vehicle's perception system may generate data regarding objects in the vehicle's environment using, for example, sensor data from one or more sensors disposed on the vehicle. In examples, the sensory data may be determined from data other than the sensor data received at operations 402, 406 described above.

[0103] At operation 504, process 500 includes identifying an object proximate to the vehicle from the sensory data. For example, the sensory data may include a representation of an object at a sensed position within the vehicle's environment. In at least some examples, operation 504 may include identifying an object within a threshold distance (e.g., 10 centimeters, 0.5 meters, 1 meter, etc.) of the vehicle.

[0104] At operation 506, process 500 includes receiving sensor data. For example, the sensor data may be sensor data 122 generated by first sensor 110 and used to determine a low energy impact collision with the vehicle, as described above in connection with FIG. 1. Also, in examples, the sensor data may correspond to the sensor data received at operation 402 and / or the sensor data received at operation 406 described above. Without limitation, the sensor data may be from a single modality, e.g., an electromechanical sensor, an acoustic sensor, etc., or may be from multiple modalities.

[0105] At operation 508, process 500 includes determining an estimated crash location from the sensor data. For example, operation 508 may correspond to operation 404 and / or operation 408 of process 400 described above. For example, operation 508 may include determining a first estimated location based on data generated by a first sensor modality and / or determining a second estimated location based on data generated by a second sensor modality. In other examples, the estimated location may be based on sensor data from one or more sensor modalities.

[0106] At operation 510, process 500 includes determining whether the estimated location corresponds to an object location. For example, operation 510 may include determining whether the object location is within a threshold distance of the estimated location. In other examples, operation 510 may include determining whether the estimated location is on a body panel or within an area where an object is in close proximity.

[0107] If, in operation 510, it is determined that the estimated position corresponds to the object position, then the process determines a crash event in operation 512. For example, operation 512 may correspond to operation 412 described above. Although not shown in FIGURE 5, process 500 may also include controlling the vehicle based on the crash event, for example, as in operation 420 described above.

[0108] Alternatively, if in operation 510 it is determined that the estimated position does not correspond to the object position, then in operation 514, process 500 includes determining a non-collision event. For example, operation 514 may correspond to operation 414 described above.

[0109] Based on the foregoing, it should be understood that while the subject matter presented herein has been described in language specific to structural components, methodological acts, computer-readable media, and / or other structural components operatively connected to the example vehicle systems 100, 200, 302, the invention defined in the appended claims is not necessarily limited to the particular features, acts, or media described herein. Rather, the particular features, acts, and media are disclosed as exemplary forms of implementing the subject matter recited in the claims.

[0110] (Example clauses) A: Electromechanical sensors; An acoustic sensor; one or more processors; one or more computer readable media storing instructions; An autonomous vehicle comprising: The instructions, when executed, cause one or more processors to: receiving electromechanical sensor data from an electromechanical sensor indicating a first crash associated with the autonomous vehicle; determining a first location of a first impact based at least in part on the electromechanical sensor data; receiving acoustic sensor data from the acoustic sensor indicating a second crash associated with the autonomous vehicle; determining a second location of a second impact based at least in part on the acoustic sensor data; and determining a crash event associated with the autonomous vehicle based at least in part on the first location corresponding to the second location; Controlling the autonomous vehicle based at least in part on the crash event; and The autonomous vehicle is caused to perform an action including:

[0111] B: the electromechanical sensor includes an impact sensor coupled to a body panel of the autonomous vehicle, the impact sensor configured to generate a signal corresponding to a magnitude of the impact; The autonomous vehicle of example A, wherein determining the first location is based at least in part on a signal having a magnitude equal to or greater than a threshold.

[0112] C: the acoustic sensor includes a first microphone coupled to the autonomous vehicle and a second microphone coupled to the autonomous vehicle and spaced apart from the first microphone; The autonomous vehicle of example A or example B, wherein determining the second location is based at least in part on analyzing a first audio signal corresponding to sound received at a first microphone and a second audio signal corresponding to sound received at a second microphone.

[0113] D: A perception system configured to generate perception data associated with objects in an environment of the autonomous vehicle, the operation further comprising: detecting an object proximate to at least one of the first location and the second location based at least in part on the sensory data; The autonomous vehicle of any one of Examples A to C, wherein determining the collision event is based at least in part on detecting an object.

[0114] E: The action in question is and controlling an autonomous vehicle based at least in part on a crash event, the controlling of the autonomous vehicle comprising: Controlling an autonomous vehicle to stop; communicating with a computing device associated with at least one of a passenger and a teleoperator associated with the autonomous vehicle; retrieving additional sensor data associated with at least one of the first estimated location and the second estimated location; generating collision event data; At least one of the following: The autonomous vehicle of any one of Examples A to D, further comprising:

[0115] F. Receiving first data from a first system associated with the vehicle; receiving second data from a second system associated with the vehicle; determining a first estimated crash location based at least in part on the first data; determining a second estimated crash location based at least in part on the second data; and determining a crash event associated with the vehicle based at least in part on the first estimated crash location corresponding to the second estimated crash location; The method includes:

[0116] G. receiving the first data includes receiving first sensor data from a plurality of electromechanical sensors; or receiving the second data includes receiving second sensor data from a plurality of acoustic sensors; The method of Example F, comprising at least one of:

[0117] H: The method of Example F or Example G, wherein the plurality of electromechanical sensors includes one or more of an impact sensor coupled to a body panel of the vehicle, a thin film sensor embedded in the vehicle, and a conductive element associated with the dashboard of the vehicle.

[0118] I: A plurality of electromechanical sensors are arranged around the vehicle, and determining a first estimated crash location includes: receiving first impact sensor data generated by a first impact sensor of the plurality of impact sensors; receiving second impact sensor data generated by a second impact sensor of the plurality of impact sensors; determining a first estimated crash location based on the first impact sensor data, the second impact sensor data, a first position of the first impact sensor in the array, and a second position of the second impact sensor in the array; The method of any one of Examples F to H, comprising:

[0119] J: The method of any one of Examples F to I, wherein receiving the second sensor data includes receiving the second sensor data from a plurality of acoustic sensors coupled to the vehicle.

[0120] K. Determining a second estimated collision location Receiving first acoustic data generated by a first acoustic sensor of the plurality of acoustic sensors; receiving second acoustic data generated by a second acoustic sensor of the plurality of acoustic sensors; determining a first estimated crash location based on the first acoustic data and the second acoustic data; The method of any one of Examples F to J, comprising:

[0121] L: The method of any one of Examples F to K, wherein receiving the first data includes receiving perception data from a perception system associated with the vehicle, the perception data including information about objects near the vehicle.

[0122] M. The method of any one of Examples F-L, including determining that the first estimated crash location is within a threshold distance from the second estimated crash location.

[0123] N: the first data includes a plurality of signal values, and determining a first estimated crash location includes: determining a cluster of sensors associated with the first system having a higher mean value using k-means; determining a first estimated impact location as a spatial center of a cluster of sensors; The method of any one of Examples F to M, comprising:

[0124] O: a first system associated with a vehicle, the first system configured to generate first data associated with the vehicle; a second system associated with the vehicle, the second system configured to generate sensor data associated with the vehicle; one or more processors; one or more computer readable media storing instructions; An exemplary vehicle comprising: The instructions, when executed, cause one or more processors to: Receiving first data; determining a first estimated crash location based at least in part on the first data; Receiving second data; and determining a second estimated crash location based at least in part on the second sensor data; determining crash data associated with the vehicle based at least in part on the first estimated position corresponding to the second estimated position; controlling the vehicle based at least in part on the crash data; A vehicle that performs an operation including:

[0125] P: receiving the first data includes receiving sensor data from a sensor; or Receiving the first data includes receiving sensory data from a sensory system. The vehicle of example O, including at least one of:

[0126] Q: A vehicle as described in any one of Examples O to P, wherein determining the first estimated location includes receiving as first data first impact sensor data generated by a first impact sensor of the plurality of impact sensors, receiving second impact sensor data generated by a second impact sensor of the plurality of impact sensors, and determining the first estimated location based on the first impact sensor data, the second impact sensor data, a first position of the first impact sensor in the array, and a second position of the second impact sensor in the array.

[0127] R: the one or more sensors include a plurality of acoustic sensors arranged in an array near a surface of the vehicle, and determining the estimated position includes: Receiving first acoustic data generated by a first acoustic sensor of the plurality of acoustic sensors; receiving second acoustic data generated by a second acoustic sensor of the plurality of acoustic sensors; determining a first estimated location based on the first acoustic data, the second acoustic data, a first position of the first acoustic sensor in the array, and a second position of the second acoustic sensor in the array; A vehicle as described in any one of Examples O to Q, including:

[0128] the one or more sensors include a plurality of electromechanical sensors and a plurality of acoustic sensors; The vehicle of any one of Examples O through R, wherein the estimated location is based at least in part on locations of a plurality of electromechanical sensors and a plurality of acoustic sensors on the vehicle.

[0129] T: A vehicle as described in any one of Examples O to S, wherein controlling the vehicle includes at least one of stopping the vehicle, transmitting operating data, and enabling two-way communication between a remote operator and a communication unit associated with the vehicle.

[0130] The subject matter described above is provided for illustrative purposes only and should not be construed as limiting. The subject matter described in the claims is not limited to implementations that solve some or all of the shortcomings described in any part of this disclosure. Various modifications and changes can be made to the subject matter described herein without following the examples and applications shown and described and without departing from the spirit and scope of the invention described in the claims.

Claims

1. An electromechanical sensor; An acoustic sensor; one or more processors; one or more computer readable media storing instructions; An autonomous vehicle comprising: The instructions, when executed, cause the one or more processors to: receiving electromechanical sensor data from the electromechanical sensor indicative of a first crash associated with the autonomous vehicle; determining a first location of the first crash based at least in part on the electro-mechanical sensor data; and receiving acoustic sensor data from the acoustic sensor indicative of a second crash associated with the autonomous vehicle; determining a second location of a second crash based at least in part on the acoustic sensor data; and determining a crash event associated with the autonomous vehicle based at least in part on the first location corresponding to the second location; controlling the autonomous vehicle based at least in part on the crash event; The autonomous vehicle is caused to perform an action including:

2. the electro-mechanical sensor includes an impact sensor coupled to a body panel of the autonomous vehicle, the impact sensor configured to generate a signal corresponding to a magnitude of an impact; determining the first location is based at least in part on the signal having a magnitude equal to or greater than a threshold. The autonomous vehicle of claim 1 .

3. the acoustic sensor includes a first microphone coupled to the autonomous vehicle and a second microphone coupled to the autonomous vehicle and spaced apart from the first microphone; determining the second location is based at least in part on analyzing a first audio signal corresponding to sound received at the first microphone and a second audio signal corresponding to sound received at the second microphone.

3. An autonomous vehicle as claimed in claim 1 or claim 2.

4. a perception system configured to generate perception data associated with objects within an environment of the autonomous vehicle; The operation includes: detecting an object proximate to at least one of the first location and the second location based at least in part on the sensory data; determining the collision event is based at least in part on detecting the object; The autonomous vehicle of claim 1 .

5. The operation includes: Controlling the autonomous vehicle based at least in part on the crash event, wherein controlling the autonomous vehicle includes: Controlling the autonomous vehicle to a stop; communicating with a computing device associated with at least one of a passenger and a teleoperator associated with the autonomous vehicle; retrieving additional sensor data associated with at least one of the first location and the second location; generating crash event data; The autonomous vehicle of claim 1 , further comprising at least one of:

6. Receiving first data from a first system associated with the vehicle; receiving second data from a second system associated with the vehicle; determining a first estimated crash location based at least in part on the first data; determining a second estimated crash location based at least in part on the second data; and determining a crash event associated with the vehicle based at least in part on the first estimated crash location corresponding to the second estimated crash location; The method includes:

7. receiving the first data includes receiving first sensor data from a plurality of electromechanical sensors; receiving the second data includes receiving second sensor data from a plurality of acoustic sensors; The method of claim 6, comprising at least one of:

8. The plurality of electromechanical sensors include an impact sensor coupled to a body panel of the vehicle; a thin film sensor embedded in the vehicle; a conductive element associated with a dashboard of the vehicle; The method of claim 7, comprising one or more of the following:

9. A plurality of electromechanical sensors are arranged in an array about the vehicle, and determining the first estimated crash location includes: receiving first shock sensor data generated by a first shock sensor of the plurality of electro-mechanical sensors; receiving second shock sensor data generated by a second shock sensor of the plurality of electro-mechanical sensors; determining the first estimated crash location based on the first impact sensor data, the second impact sensor data, a first position of the first impact sensor in the array, and a second position of the second impact sensor in the array; The method of claim 8 , comprising:

10. The method of claim 7 , wherein receiving the second sensor data comprises receiving the second sensor data from a plurality of acoustic sensors coupled to the vehicle.

11. Determining the second estimated crash location includes: Receiving first acoustic sensor data generated by a first acoustic sensor of the plurality of acoustic sensors; receiving second acoustic sensor data generated by a second acoustic sensor of the plurality of acoustic sensors; determining the first estimated crash location based on the first acoustic sensor data and the second acoustic sensor data; The method of claim 10, comprising:

12. The method of claim 6 , wherein receiving the first data comprises receiving sensory data from a perception system associated with the vehicle, the sensory data comprising information about objects proximate to the vehicle.

13. The method of claim 6 comprising determining that the first estimated crash location is within a threshold distance from the second estimated crash location.

14. The first data includes a plurality of signal values, and determining the first estimated crash location includes: determining a cluster of sensors associated with the first system having a higher mean value using k-means; determining the first estimated crash location as a spatial center of a cluster of the sensors; The method of claim 6, comprising:

15. 7. A non-transitory computer-readable medium storing instructions executable by one or more processors to perform the method of claim 6.

Images