A smart system for determining a driver's driving behavior

Abstract

A system (1) for efficiently determining the driving behavior of a driver operating a vehicle. The system includes a three-axis accelerometer (2) for generating accelerometer data (3), an orientation sensor (4) for generating orientation data (5), and an attitude sensor (6) for generating attitude and angular velocity data (7), all located within the vehicle. The system (1) also includes a processing unit (8) for processing the accelerometer data (3), the orientation data (5), and the attitude data (7) and calibrating the accelerometer data (3) to generate calibrated accelerometer data (9). The calibrated accelerometer data (9) is used to determine the driving behavior of the driver of the vehicle and generate driving error alerts, including, but not limited to, speeding, hard acceleration, hard braking, fast cornering, abrupt lane changing, distracted driving, dangerously driving over bumps, and crash detection.

Description

[Technical field]

[0001] The present invention relates to a driving behavior monitoring system, more specifically, to the driving behavior of a driver operating a vehicle by calibrating, correcting and analyzing data obtained from sensors during driving. [Background technology]

[0002] In modern times, with the increase in road accidents, road safety has become an increasing concern for everyone. Driving behavior and road accidents are directly related to each other. Around 1.35 million people die in road accidents every year worldwide. Most can be avoided by following certain safety measures. However, in today's hectic world, these measures are often overlooked by drivers. Existing methods of monitoring the driving behavior of drivers operating vehicles include using standalone telematics data processing systems, driving guidelines, training, etc. to overcome accident risks and safe driving practices.

[0003] That method involves using telematics devices in the vehicle, which is expensive and involves the additional logistics of installing the devices in the vehicle. Furthermore, traditional smartphone-based driving behavior monitoring solutions are not accurate.

[0004] In light of the foregoing discussion, there is a need to develop an economical and accurate solution for safe driving practices to reduce accident risks, which provides highly accurate driving behavior monitoring without the use of any dedicated telematics device in the vehicle. There is also a need for a solution that can be used to rank drivers based on their safe driving scores and reward them for safe driving, specifically as part of a fleet management system or usage-based insurance program to ensure safe driving practices and reduce accident risks.

[0005] It is an object of the present invention to provide a system for determining the behavior of a driver driving a vehicle.

[0006] Another object of the present invention is to provide a system that can accurately and robustly determine driving behavior using data from an in-vehicle mobile communication device, such as the driver's smartphone. Optionally, the mobile communication device is communicatively paired with a low-cost Inertial Measurement Unit (IMU) sensor, further improving the accuracy of driving behavior while minimizing added cost and inconvenience to the driver.

[0007] Another object of the present invention is to provide a system for calibrating accelerometer data of a vehicle and preparing the data for determining the driving behavior of a driver operating the vehicle, which may be a vehicle with four or more wheels (such as cars, trucks and buses) as well as two-wheeled vehicles (such as motorcycles and bicycles).

[0008] It is a further object of the present invention to provide a system that combines different rash and risky driving behaviors to evaluate a driver for safe driving. Summary of the Invention

[0009] A system is provided for efficiently determining the driving behavior of a driver operating a vehicle. The system includes a three-axis accelerometer for generating acceleration data, an orientation sensor for generating orientation data, and an attitude sensor for generating attitude data. A processing unit of the system processes the data through a series of processing steps to generate accelerometer data calibrated such that the three axes of the accelerometer are aligned with the orientation of the vehicle. The system uses the calibrated accelerometer data in conjunction with the orientation data to determine various inappropriate driving behavior occurrences using a series of processing steps. The system is adapted to generate poor driving alerts including, but not limited to, speeding, rapid acceleration, rapid braking, rapid cornering, rapid lane changing, distracted driving, dangerously driving over bumps, and collision detection. The system is adapted to determine the driving behavior of vehicles with four or more wheels (such as cars, trucks, buses, etc.) as well as two-wheeled vehicles that can lean sideways (such as motorcycles, bicycles, etc.). The system is adapted to determine the driving behavior using data from a mobile communication device in the vehicle, such as the driver's smartphone. Optionally, the mobile communications device is communicatively paired with a low-cost Inertial Measurement Unit (IMU) sensor to further improve accuracy of the driving behavior. [Brief description of the drawings]

[0010] In this description, novel features and characteristics of the present disclosure are set forth. However, the disclosure itself, as well as its preferred modes of use, further objects and advantages, will best be understood by reference to the following description of exemplary embodiments, when read in conjunction with the accompanying drawings, in which: One or more embodiments will now be described, by way of example only, with reference to the accompanying drawings in which like reference numerals represent like elements, and in which:

[0011] [Figure 1] 1 shows a schematic diagram of a system for calibrating accelerometer data of a vehicle. [Figure 2-1] 1 illustrates another embodiment of a portion of a system associated with generating calibrated acceleration data. [Figure 2-2]1 illustrates a sequence of processing steps performed within a processing system associated with generating calibrated acceleration data. [Figure 2-2 (continued)] 1 illustrates a sequence of processing steps performed within a processing system associated with generating calibrated acceleration data. [Figure 3-1] Illustrated are portions of a system for determining various inappropriate driving behavior occurrences. [Figure 3-2] 1 illustrates a sequence of processing steps in a system for determining various inappropriate driving behavior occurrences. [Figure 3-2 (continued)] 1 illustrates a sequence of processing steps in a system for determining various inappropriate driving behavior occurrences. [Figure 3-2 (continued2)] 1 illustrates a sequence of processing steps in a system for determining various inappropriate driving behavior occurrences. [Figure 4] 1 illustrates a portion of a system for determining various improper driving behavior events of a vehicle leaning to the side. [Diagram 5] FIG. 1 shows a schematic diagram of another embodiment of a system for driving behavior monitoring and feedback, including data capture, data calibration, generation of inappropriate driving behavior alerts and safe driving, and a user interface for a driver.

[0012] These figures depict embodiments of the present disclosure for purposes of example only. Those skilled in the art will readily recognize from the following description that alternative embodiments of the assemblies, structures, and methods illustrated herein may be employed without departing from the principles of the present disclosure as described herein. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] For the purposes of promoting an understanding of the principles of the invention, reference will now be made to the embodiments illustrated in the drawings, and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the invention is intended thereby. Such changes and further modifications in the illustrated system, and such further applications of the principles of the invention normally occurring to those skilled in the art, are to be construed as being within the scope of the invention.

[0014] It will be understood by those skilled in the art that the foregoing summary and the following detailed description are exemplary and explanatory of the invention and are not intended to be limiting thereof.

[0015] The terms "comprises," "comprising," or any other variations thereof, are intended to encompass a non-exclusive inclusion, such that a process or method that includes a list of steps does not include only those steps, but may include other steps not expressly recited in or inherent to such process or method. Similarly, one or more subsystems, or elements or structures or components described as "comprises...a" does not, without further constraints, preclude the presence of other subsystems, elements, structures, components, additional subsystems, additional elements, additional structures, or additional components. The appearances of phrases such as "in one embodiment," "in another embodiment," and similar grammatical expressions throughout this specification may, but do not necessarily, all refer to the same embodiment.

[0016] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The systems, methods, and examples provided herein are illustrative only and are not intended to be limiting.

[0017] The systems, methods, and examples provided herein are exemplary only and are not intended to be limiting.

[0018] Embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0019] The present disclosure relates to monitoring the driving behavior of a driver operating a vehicle using data from the driver's smartphone. Optionally, a low-cost inertial measurement unit (IMU) is installed in the vehicle that is combined with the driver's smartphone data. The solution also generates a composite and normalized safe driving score by combining various driving behavior parameters.

[0020] FIG. 1 shows a schematic diagram of a system (1) for calibrating accelerometer data of a vehicle and for determining the driving behavior of a driver operating the vehicle. The system (1) includes a three-axis accelerometer (2), a direction sensor (4), and an attitude sensor (6) installed in the vehicle. The three-axis accelerometer (2) generates acceleration data (3) related to the acceleration of the accelerometer at each time instant. The direction sensor (4) generates direction data (5) related to the direction in which the three-axis accelerometer (2) is traveling. The attitude sensor (6) generates attitude data (7) related to the attitude and / or angular velocity of the three-axis accelerometer (2).

[0021] The system 1 also includes a processing unit 8 that processes the accelerometer data 3, the orientation data 5, and the attitude data 7, recalibrates the accelerometer data 3, and generates calibrated accelerometer data 9. The processing unit 8 recalibrates the accelerometer data 3 such that three axes of the accelerometer data 3 match three axes of the vehicle attitude, such that the calibrated accelerometer data 9 is useful for determining inappropriate driving behavior events.

[0022] The driving behavior events considered for analysis and determination may include speeding, sudden acceleration, sudden braking, fast cornering, sudden lane changing, fatigued driving, collision detection, and distracted driving, etc. These events need to be captured and analyzed as part of driver behavior monitoring.

[0023] These data (3, 5, 7) may be received from sensors (2, 4, 6) in a mobile communication device carried by the driver, or from sensors (2, 4, 6) in a separate IMU installed in the vehicle, as well as some combination of both. The mobile communication device may have a 3-axis accelerometer (2), a direction sensor (4), and an attitude sensor (6). The direction sensor (4) may be a GPS / GNSS sensor that can be used to obtain direction and speed data. The attitude sensor may be a gyroscope sensor that produces angular rate data. The IMU unit may have a 3-axis accelerometer (2) and an attitude sensor (6) to provide accelerometer and attitude data, respectively. The IMU unit may be battery powered. The IMU unit may transmit the sensor data to the driver's mobile communication device, which may also be the driver's smartphone, using a Bluetooth low energy connection.

[0024] It should be noted that the mobile communication device is generally picked up by the driver multiple times during the commute, which results in changes in the smartphone's attitude. This means that the accelerometer data calibration system (1) takes such attitude changes into account so that the three axes of the accelerometer data are correctly aligned with the vehicle's attitude even after each such change in smartphone attitude. When receiving data (3,5,7) from an IMU unit fixed in the vehicle, calibration is only required once at the time of initial installation, and at most when removed for any repair. Thus, in an embodiment in which data from the IMU unit is used, the computational requirements of the system (1) are reduced while calculating the driver's behavior. Such an embodiment also ensures that data is captured from the designated vehicle in which the IMU is installed, thus improving the accuracy of the determination of the driver's behavior.

[0025] Furthermore, each time the driver starts driving the vehicle, the processing unit (8) may automatically start receiving data from the sensors (2, 4, 6) for further determination of the driving behavior.

[0026] The advantage of using an IMU unit is that even if the smartphone’s GPS / GNSS sensors are switched off (to save the smartphone’s battery), the sensor data received from the IMU unit still allows the system (1) to identify certain important driving behavior events (alerts) such as hard acceleration, hard braking, fast cornering, sudden lane changes, etc.

[0027] After calibrating the accelerometer data (3), the processing unit (8) compares the corrected accelerometer data (9) in the vehicle's XY plane at each time point with a fifth predefined threshold (not shown in the figure). If the magnitude of the corrected accelerometer data (9) exceeds the fifth predefined threshold, the processing unit (8) determines such time point as an inappropriate driving behavior event (30).

[0028] Figures 2-1 and 2-2 are taken together to provide a more detailed description of generating the calibrated acceleration data 9. Figure 2-1 illustrates another embodiment of a portion of the system 1 associated with generating the calibrated acceleration data 9, while Figure 2-2 illustrates a series of processing steps taking place within the processing unit 8 associated with generating the calibrated acceleration data 9.

[0029] In this embodiment, the processing unit (8) uses the acceleration data (3), orientation data (5), and attitude data (7) to generate angle data relating to the three rotation angles (i.e., φ, θ, and ψ corresponding to the Euler ZYZ transformation) required to rotate the Z, Y, and Z axes of the accelerometer in turn at each instant in time to align the accelerometer with the three axes of the vehicle's attitude at each instant in time to generate calibrated accelerometer data (9).

[0030] To further improve the performance of the calibration of the accelerometer data (3), only the attitude data (7) representing data from consecutive events where the difference in the attitude data (7) is within a first predefined threshold (12) is considered for the calibration. To this end, the processing unit (8) receives the attitude data (7) for a predefined time and then checks the difference in the consecutive attitude data (7) for various events within the predefined time against the first predefined threshold (12). If the difference in the attitude data (7) for consecutive events or a consecutive series of events is within the first predefined threshold (12), those sections of the events are labeled as stable attitude data (10). And if the difference in the attitude data (7) of consecutive events is outside the first predefined threshold (12), they are labeled as unstable attitude data (11). The processing unit (8) uses the stable attitude data (10) in conjunction with the accelerometer data (3) to recalibrate the accelerometer data (3) and generate calibrated accelerometer data (9).

[0031] To further improve the performance of the calibration of the accelerometer data (3), the processing unit (8) processes the stabilized attitude data (10) in the following manner: The processing unit (8) processes the stabilized attitude data (10) to generate angle data (18). The angle data (18) includes a prior φ angle (19, representing a first rotation angle about the Z axis of the reference frame) of the attitude of the three-axis accelerometer (2) relative to the f frame of the vehicle, and a θ angle (20, representing a tilt about the Y axis of the reference frame) of the attitude of the three-axis accelerometer (2) relative to the f frame of the vehicle. The processing unit (8) compares the change (21) of the angle data with a second predetermined threshold (22). If the change (21) of the angle data is within the second predetermined threshold (22), the processing unit (8) divides the angle data (18) into smaller subsection periods and calculates a moving average (23) of the angle data. If the change in the angle data (21) is within a second predetermined threshold (22), the processing unit (8) generates a median value (24) of the angle data. The processing unit (8) processes the moving average (23) of the angle data or the median value (24) of the angle data together with the accelerometer data (2) to recalibrate the accelerometer data (2) and generate intermediate calibrated accelerometer data (9-1) in the plane of the vehicle.

[0032] To further improve the performance of the calibration of the accelerometer data (3), the processing unit (8) compares the intermediate calibrated accelerometer data (9-1) in the plane of the vehicle with a third predefined threshold (25) and simultaneously compares the orientation data (5) for the corresponding period with a fourth predefined threshold (26). If the corrected accelerometer (9) is above the third predefined threshold (25) and the orientation data (5) is below the fourth predefined threshold (26), the processing unit (8) determines the ψ angle (27) of the attitude of the three-axis accelerometer (2) relative to the reference frame of the vehicle for the stable period. The ψ angle (27) of the attitude of the three-axis accelerometer (2) represents a second rotation around the Z axis of the reference frame. The processing unit (8) further determines the most frequently occurring or most likely ψ angle (28). The processing unit then processes the angle data (18) together with the most likely ψ angle (28) and the accelerometer data (3) to recalibrate the accelerometer data (3) and generate the final calibrated accelerometer data (9). In one embodiment of the present invention, a ψ value can be selected from multiple ψ values ​​using a kernel density estimation method for a probability density distribution, and then the most likely ψ can be found.

[0033] If the accelerometer data (2) is captured from the driver's smartphone, the smartphone's attitude within the vehicle may change whenever the driver picks up or moves the smartphone while driving. This means that the accelerometer data (2) needs to be recalibrated so that its three axes match the attitude of the vehicle. The GPS / GNSS sensor and the gyroscope sensor within the smartphone provide the orientation data (5) and attitude data (7), respectively. The above methodology is applied to situations such as those detailed herein below. 1. The purpose of smartphone attitude correction is to transform the smartphone's three-dimensional reference frame into the vehicle's reference frame. That is, the acceleration data recorded by the accelerometer in three dimensions (x,y,z) is transformed into corresponding acceleration data in the vehicle's direction (X,Y,Z). This corrected acceleration data can then be used for the remaining steps of sensor data processing. The following steps describe the transformation of acceleration data from the smartphone's reference frame into the vehicle's reference frame. 2. The ZYZ Euler angle transformation is used to describe the three angles needed to change the attitude: φ (phi, the first rotation about Z), θ (theta, the tilt about Y), and ψ (psi, the second rotation about Z). 3. The sensor data for the entire ride is backed up during periods when the gyroscope is relatively stable. Anywhere the gyroscope data is relatively unstable indicates movement of the phone due to some user action. 4. For the stable period, the smoothed accelerometer readings are used to calculate the φ (first rotation about Z) and θ (tilt about Y) angles at each time using trigonometric equations based on specific Euler angles. 5. If the θ and φ angles are found to change more than a certain threshold within the stable period, this indicates that the phone's position is changing due to some natural drift or slippage of a moving vehicle. In such a case, the processing unit further divides the stable period into smaller sub-section periods and calculates the running averages of φ and θ. 6. If within the stable period the processing unit finds that the theta and phi angles have not changed by more than a certain threshold, the processing unit uses the median acceleration value to calculate a single theta and phi angle for the entire stable value. 7. Now, for each stable period or subsection, the processing unit finds the corrected accelerometer readings using the φ and θ angles by applying trigonometric equations based on the specific Euler angles. The processing unit finds the corrected accelerometer readings where the magnitude of the acceleration in the vehicle's XY plane (x2+y2) is above a specific threshold and the change in GPS course around that time is below a specific threshold. These are events of interest where hard braking or acceleration may have occurred. 8. For each such event, the processing unit calculates the ψ angle using the corresponding φ and θ values ​​and the accelerometer data for that time, and applies them to a trigonometric equation based on the particular Euler angles. For each stable period or subsection, the processing unit generates multiple ψ values, and the ψ value with the highest frequency or likelihood is selected that matches the direction of the GPS velocity change for that event. Steps 7 and 8 above ensure that the ψ angle determination is good even if the particular GPS data is noisy or inaccurate, since the angle is calculated based on the majority view from the different events (both acceleration and deceleration) collected for each stable period or subsection. This improves the robustness of smartphone attitude correction compared to existing techniques. 9. At the end of this process, the processing unit generates three angles, φ, θ, and ψ, for each stable period or subsection. These are the three angles used to transform the accelerometer attitude readings from the smartphone's frame of reference to the vehicle's frame of reference. 10. For any stretch of driving, if the processing unit cannot conclusively determine the three transformation angles, the GPS sensor data is only used to determine the driving behavior parameters for that stretch of driving, which further improves the robustness of the system.

[0034] For fixed IMU units, the accelerometer data from the IMU unit needs to be calibrated so that its three axes are aligned with the vehicle's orientation. However, this calibration only needs to be done once and is simple compared to how smartphone sensor data is calibrated. The steps for this one-time calibration are as follows: 1. The objective is to convert the acceleration data in three dimensions (x,y,z) recorded by the accelerometers in the IMU sensor into corresponding acceleration data in the vehicle's direction (X,Y,Z). 2. The processing unit uses the ZYZ Euler angle transformation to describe the three angles needed to change the pose: φ (first rotation around Z), θ (tilt around Y), and ψ (second rotation around Z). 3. Sensor data is collected from one trip of a vehicle equipped with an IMU unit. The processing unit calculates the φ angle (first rotation around Z) and theta angle (tilt around Y) at each time instant using trigonometric equations based on specific Euler angles. The processing unit takes the median values ​​of the φ and θ angles as the fixed φ and θ angles to use for the conversion. 4. The processing unit determines the corrected accelerometer data using the φ and θ angles calculated in step 3 above by applying trigonometric equations based on certain Euler angles. The processing unit finds corrected accelerometer data that are above a certain threshold and the change in GPS orientation around that time is below a certain threshold. These are events of interest where hard braking or acceleration may have occurred. 5. For each such event, the processing unit calculates a posterior ψ angle using the corresponding prior φ and θ values ​​and the accelerometer value at that time, and applies them to a trigonometric equation based on the particular Euler angles. For each stable period or subsection, the processing unit determines multiple posterior ψ values, and the ψ value with the highest frequency or likelihood is selected to match the direction of the change in GPS velocity during that event. 6. At the end of this process, the processing unit generates three angles, namely φ, θ, and ψ, based on the accelerometer data collected from one run. 7. Because the IMU units are fixed, these are fixed angles that the processing unit uses to correct the accelerometer data of any future runs without having to recalculate φ (first rotation about Z), θ (tilt about Y), and ψ (subsequent rotations about Z). 8. This saves on the computational resource requirements of the system. Also, in the case of a fixed IMU unit, even if the GPS sensor of the user's smartphone is switched off for the rest of the journey (to save battery), the system can still generate driving alerts for sudden acceleration, hard braking, fast cornering, sudden lane changes, etc.

[0035] Figures 3-1 and 3-2 are taken together to provide a detailed description of determining various inappropriate driving behavior events. Figure 3-1 illustrates a portion of the system (1) for determining various inappropriate driving behavior events, namely, a sudden acceleration or deceleration section (33), a fast cornering or sudden lane change section (36), an excessive speed section (44), a distracted driving time section (48), a dangerous bump driving event (52), and a collision detection event (55). Meanwhile, Figure 3-2 illustrates a sequence of processing steps in the system for determining various inappropriate driving behavior events as described below.

[0036] The processing unit (8) is adapted to compare the corrected accelerometer data (9) in the horizontal (XY) plane of the vehicle at each time point with a fifth predetermined threshold (29), and if the magnitude of the corrected accelerometer data (9) exceeds the fifth predetermined threshold (29), the processing unit (8) is adapted to determine such time point as an inappropriate driving behavior event (30).

[0037] The processing unit (8) determines a first pattern (31) within the accelerometer data (9) corrected for the vehicle's direction of travel for consecutive time intervals of various inappropriate driving behavior occurrences (30). If the first pattern (31) matches a first predefined pattern (32), the processing unit determines the time interval as an interval of rapid acceleration or deceleration (33), which is useful for determining sudden braking and rapid acceleration events.

[0038] The processing unit (8) determines a second pattern (34) within the corrected accelerometer data (9) in a direction perpendicular to the direction of vehicle motion for consecutive time intervals of various inappropriate driving behavior events (30). If the second pattern (34) exceeds a second predetermined pattern (35), the processing unit (8) determines the time interval as a high speed cornering or abrupt lane change interval (36). This helps to determine high speed cornering and abrupt lane change type events.

[0039] The pattern matching technique used can classify the corrected accelerometer data for each time into relevant categories based on a dynamic time warping (DTW) based pattern matching technique. This technique uses a sliding window to match the corrected accelerometer pattern with a reference DTW pattern. A sliding window is a collection of corrected accelerometer readings around a center point. For each point, the sliding window is shifted by a certain number of points to match the DTW pattern. Consecutive sliding windows that match the DTW pattern associated with a driving alert are combined to label one time of consecutive driving alerts as a single event. This allows for high accuracy identification of driving alerts while avoiding double counting of driving alerts.

[0040] It should be noted that the processing unit 8 does not use position data from the GPS / GNSS sensor to determine sections of rapid acceleration or deceleration 33 and sections of rapid cornering or lane changes 36. Thus, in embodiments where a fixed IMU unit sensor is used and correction of the accelerometer data 2 is only required once (and not for each trip), the processing unit 8 can generate driving alerts even if the GPS / GNSS sensor of the mobile communication device is switched off for battery saving purposes.

[0041] For sections of driving where accelerometer data (2) is not available or where corrected accelerometer data (9) cannot be determined, the GPS sensor data can be used to determine sections of rapid acceleration or deceleration (33), and sections of rapid cornering or rapid lane changes (36). Vehicle speed is calculated from the GPS sensor data by differentiating the distance between two successive GPS readings with the time difference between the two successive readings. This vehicle speed at each time instant is further used to calculate the acceleration and deceleration of the vehicle while it is moving.

[0042] For high speed cornering detection or sudden lane change segments (36), the GPS orientation data (5) is used to calculate the angular velocity, and the speed data obtained from the GPS is used to define the tangential velocity. The two are multiplied to calculate the centripetal acceleration, which is indicative of the cornering force. If the cornering force is above a certain threshold, it is considered a high speed cornering or sudden lane change.

[0043] The processing unit (8) processes sections of rapid acceleration or deceleration (33) and sections of rapid cornering or rapid lane changes (36), and if sections of the same type (33, 36) are within a predetermined period (37), the processing unit (8) combines sections of the same type and generates a driving problem warning (38) of that type for the combined sections.

[0044] Additionally, the processing unit (8) receives speed change information (39) and direction change information (40) for the various improper driving behavior events (30), correlates the poor driving alert (38) with the speed change information (39) and direction change information (40), and determines the accuracy and severity (41) of the poor driving alert (38). In one embodiment, the processing unit (8) may use either the speed change information (39) or the direction change information (40) to determine the accuracy and severity (41) of the poor driving alert (38).

[0045] The processing unit (8) processes the corrected accelerometer data (9) and the direction data (5) for various improper driving behavior events (30) to determine the vehicle's speed (42) during those events. In one embodiment, the processing unit (8) uses a filter, such as a Kalman or complementary filter, to determine the vehicle's speed (42). The processing unit (8) compares the vehicle's speed (42) to a first predetermined threshold speed (43), and if the speed of the vehicle (42) exceeds the first predetermined threshold speed (43), the processing unit (8) determines the event as a speeding event (44). In one embodiment, the first predetermined threshold speed (43) is dynamically determined based on the vehicle's location provided by a GPS sensor and location-based road speed limit information available from an external source.

[0046] The processing unit (8) processes the posture data (7) during the period (11) of unstable posture data to determine a posture data pattern (45). The processing unit (8) further compares the posture data pattern (45) to a predefined posture pattern (46) and, if there is a match, determines the time period as a time period of phone use (47). The processing unit (8) then correlates the time period of phone use (47) with a second predefined speed threshold (43-1) and determines the time period of phone use as a time period of distracted driving (48).

[0047] In the event of a bump detection, the processing unit (8) processes the corrected accelerometer data (9) in a ground direction for a short subsection of the period (10) of stable attitude data to determine a ground direction data pattern (49) of the corrected accelerometer data. The processing unit (8) further compares the data pattern (49) to a fifth predetermined data pattern (50) and, if there is a match, determines the period as a bump detection period (51). The processing unit (8) then correlates the bump detection period (51) with a third predetermined speed threshold (43-2) and determines the bump detection period as a dangerous bump crossing event (52).

[0048] To determine a crash detection event (55), the processing unit (8) processes the vehicle's speed (42) over a short interval to determine a sudden decrease in speed (53). The processing unit (8) further processes the corrected accelerometer data (9) around the same time interval to determine an acceleration pattern (54). The processing unit (8) then correlates the sudden speed change (53) with the acceleration pattern (54) and matches the acceleration pattern (54) to a sixth set of predefined patterns (61) to determine a crash detection event (55). In one embodiment, the processing unit (8) may specifically match the acceleration pattern in the horizontal (XY) plane of the vehicle to the sixth predefined pattern set (61) to determine the direction relative to the vehicle that the crash occurred.

[0049] In one embodiment, this crash detection event (55) can be transmitted in real time to a cloud server using wireless connectivity provided by the user's mobile communications device. Also, processed data around the time of the crash alert, such as approach speed, direction, location, driving behavior, etc., is transmitted to the cloud server. Depending on usage, this information in the cloud server is used to notify emergency response services and provide roadside assistance. It can also be used for accident reconstruction and auto insurance claims management.

[0050] It should be noted that the processing unit (8) is further capable of processing all the inappropriate driving behavior events and generating a composite score or ranking for assessing the driver's driving skills.

[0051] Figure 4 illustrates a portion of a system (1) for determining various inappropriate driving behavior events for sideways leaning vehicles such as bicycles, scooters, etc. To maintain the brevity of this disclosure, only those elements having processing differences relative to the system of Figure 3 are described. The processing differences lie in determining some of the inappropriate driving behavior events (30), specifically the sections of rapid acceleration or deceleration (33) and rapid cornering (36) described below.

[0052] The processing unit (8) compares the corrected accelerometer data (9) within the three-dimensional frame of the vehicle at each time point with a ninth predetermined threshold (56). If the magnitude of the corrected accelerometer data (9) exceeds the ninth predetermined threshold (56), the processing unit (8) determines such time point as an inappropriate driving behavior event (30).

[0053] The processing unit (8) determines a third pattern (57) in the corrected accelerometer data (9) in the direction of vehicle motion for successive time intervals. If the third pattern (57) matches a third predefined pattern (58), the processing unit (8) determines the time interval as a period of rapid acceleration or deceleration (33). The processing unit (8) further determines a fourth pattern (59) in the corrected accelerometer data (9) in a plane perpendicular to the direction of vehicle motion for successive time intervals. If the fourth pattern (59) exceeds a fourth predefined pattern (60), the processing unit (8) determines the time interval as a period of rapid cornering (36). A pattern matching technique based on a dynamic time warping based pattern matching technique can be used to classify the corrected accelerometer data for each time interval into a relevant category.

[0054] FIG. 5 shows a schematic diagram of an embodiment of a system for driver behavior monitoring and feedback, including data capture, data calibration, generation of inappropriate driving behavior warnings and safe driving, and a user interface for the driver.

[0055] The system allows for the flexibility of adding new safe driving factors to be considered. It also vectorizes (parallelizes) the safe driving factor calculation across a large number of raw sensor data points to improve performance of the calculation of safe driving factors. Different thresholds are applied depending on the factors considered. The thresholds depend on the type of vehicle and the geographic location where the vehicle is driven. In general, additional criteria can be added to change the thresholds considered for each type of factor.

[0056] Further, the processing unit may be configured to assign a score for each individual factor and calculate a composite safe driving score. Each of the factors that affect safety (speeding, hard acceleration, hard braking, fast cornering, distracted driving, dangerously driving over bumps, etc.) may be assigned a score out of 100. The logic for calculating the score associated with each factor may vary based on the factor. In general, a factor may be assigned a score of 100 for a given driver / vehicle if there were no abnormal or hasty driving events for that variable in the driving. If there were hasty driving events for a variable in the driving, the number, duration, and severity of each event may be evaluated and combined to assign an overall percentage rating for that parameter for the driving.

[0057] The scores of the individual factors are combined to generate an overall safe driving score for the drive or trip. The scores across all trips can be combined by a processing unit using a weighted average of the distance traveled in each drive to give an overall safe driving score. Trends of the safe driving scores across trips and time are used to improve driving skills and promote safe driving behavior in the driver.

[0058] In some embodiments, techniques can be used to normalize the safe driving scores for different driving conditions, such as highway or non-highway, vehicle weight, daytime or nighttime driving, and general weather conditions while driving based on the location of the drive. These normalizations can allow the safe driving scores to more accurately reflect risk. For example, speeding in the rain or at night is more dangerous than speeding during the day in clear weather. These normalizations can be compared between drivers to assess and rank drivers in a group for the purpose of safe driving ability, gamification, and training.

[0059] The system 1 can utilize the sensor data received from the driver's smartphone itself where the sensor data is already aggregated. This reduces the cost and time of transferring large amounts of data to another computing resource. This also addresses the user's privacy concerns by not sending any personal information such as GPS location to the server. At the same time, the system can use the GPS location information locally to generate a driver score that can be sent to the server. The user can further view the analyzed driving behavior events including details and route of each trip in the user interface of the application running on the driver's smartphone. In another embodiment, the captured sensor data is transferred in real time to the vehicle's head unit using a connection technology such as Bluetooth or Wifi, and the sensor data is processed on the processing resources in the vehicle's head unit. In yet another embodiment of the invention, the captured sensor data is transferred to a cloud-based server using a connection technology such as cellular / Wifi in real time or on a batch basis, and the processing of the sensor data is performed on the processing resources of the cloud-based server. In another embodiment of the invention, the processing of the sensor data can be switched between running on the driver's smartphone and running on a cloud-based basis depending on the availability of resources and the connectivity between the smartphone and the cloud.

[0060] The analyzed driving behavior events from the driver's trips and the safe driving score trends are displayed to the driver using a smartphone-based mobile application running on the driver's smartphone. The app allows the driver to see the driving behavior of each trip as well as an aggregate score for all trips. The driver can analyze which components of their driving need to be improved by drilling down into a detailed alert view of each trip. The composite driving score is also used to compare and show the ranking of drivers within a group. Other applications, such as web-based dashboard applications, can be used to provide a similar user interface. The analyzed data is sent to a cloud server and can be further shared with third parties, such as insurance companies and car companies, with the driver's consent.

[0061] Regardless of whether data is captured from a mobile communication device (smartphone) or an IMU unit, the steps involved in driving behavior (alert) generation and safe driving score generation remain common, which increases the flexibility of the system and reduces the overall memory footprint of the system, allowing it to be deployed on memory-constrained mobile devices.

[0062] In another embodiment, the sensor data is processed during ongoing driving and audio or visual alerts are generated on the driver's smartphone during ongoing driving to alert the driver when driving behavior is unsafe.

[0063] It should be noted that in the case of privately owned vehicles, the system can be used to provide guidance to the vehicle owner or driver to drive safely. In the case of fleet vehicles owned by an organization, the system can be used as part of a fleet management system to ensure safe driving practices and reduce accident risks. In the case of an automobile insurance company, the system can be used to evaluate the accident risk of each driver or vehicle that the insurance company insures. The system can be used to create an underwriting model as part of a usage-based insurance policy, where the premium for the insurance policy varies depending on the driving behavior of one or more drivers who operate a particular vehicle. Vehicle manufacturers can use the system to understand how drivers operate their vehicles under various driving conditions and design vehicles with appropriate safety measures, controls, and warning systems. The present invention is also useful for various other stakeholders in the automotive ecosystem who are interested in monitoring driving behavior and providing solutions based on driving behavior.

[0064] List of reference signs 1. System 2.3-axis accelerometer 3. Accelerometer Data 4. Direction Sensor 5. Orientation Data 6. Attitude Sensor 7. Attitude Data 8. Processing Unit 9. Calibrated Accelerometer Data 9-1. Intermediately corrected accelerometer data 10. Stable attitude data 11. Unstable attitude data 12. First Predetermined Threshold 18.Angle Data 19. Attitude φ angle 20. Attitude θ angle 21. Change in angle data 22. Second Predetermined Threshold 23. Moving average of angle data 24. Median of angle data 25. Third Predetermined Threshold 26. Fourth Predetermined Threshold 27. Psi angle of attitude 28. Most frequent or most probable ψ angle 29. Fifth Predetermined Threshold 30. Inappropriate Driving Behavior Incidents 31. First pattern of corrected accelerometer data 32. First Predetermined Pattern 33. Sections of sudden acceleration or deceleration 34. Second pattern of corrected accelerometer data 35. Second Predetermined Pattern 36. Sections with high-speed cornering and sudden lane changes 37. Specified Period 38. Poor operation alarm 39. Speed ​​change information 40. Direction change information 41. Accuracy and severity of driving failure alarms 42. Vehicle speed 43. First predetermined speed threshold 43-1. Second Predetermined Speed ​​Threshold 43-2.Third Predetermined Speed ​​Threshold 44. Speeding incident 45. Posture data pattern 46. ​​Prescribed posture pattern 47. Time interval of telephone use 48. Distracted Driving Time Segments 49.Acceleration data pattern in vehicle direction 50. Fifth Predetermined Data Pattern 50-1. Sixth Predetermined Data Pattern 51. Bump detection section 52. Dangerous Bump Jumping Incident 53. Sudden changes in speed 54.Acceleration Pattern 55. Collision detection event 56. 9th Predetermined Threshold 57.Third pattern in accelerometer data corrected for vehicle orientation 58. Third Predetermined Pattern 59. Fourth pattern of accelerometer data corrected in a plane perpendicular to the vehicle 60. The Fourth Predetermined Pattern

Claims

1. A system (1) for efficiently determining a driving behavior of a driver who drives a vehicle, the system comprising: a three-axis accelerometer (2) installed in the vehicle, the three-axis accelerometer (2) being adapted to generate accelerometer data (3) relating to the acceleration of the vehicle at each instant in time; a direction sensor (4) installed in the vehicle and adapted to generate direction data (5) related to the direction in which the three-axis accelerometer (2) is traveling; an attitude sensor (6) installed in the vehicle and adapted to generate attitude data (7) related to the attitude and angular velocity of the three-axis accelerometer (2); a processing unit (8) adapted to process the accelerometer data (3), the orientation data (5) and the attitude data (7) and to recalibrate the accelerometer data (3) to generate accelerometer data (9) calibrated so that the three axes of the accelerometer coincide with the orientation of the vehicle, The system (1), wherein the calibrated accelerometer data (9) is used to determine the driving behavior of the driver of the vehicle.

2. 2. The system (1) of claim 1, wherein the processing unit (8) is adapted to process the orientation data (5) to determine an offset of the orientation data (5) and, based on such determination, to process the accelerometer data (3) and the attitude data (7) to recalibrate the accelerometer data (3) and generate the calibrated accelerometer data (9).

3. The processing unit (8) receives the attitude data (7) for a given time, collates the attitude data (7) for various events within the given time, and divides the attitude data (7) into two sections: - stable attitude data (10) for one or more sections within said predefined time period when, for a consecutive event or a consecutive sequence of events, the difference between said attitude data (7) is within a first predefined threshold (12); - for a consecutive event or a consecutive series of events, unstable attitude data (11) for one or more sections within said predetermined time, when said difference in attitude data (7) is outside said first predetermined threshold (12), 3. The system (1) of claim 2, wherein the processing unit (8) is adapted to process the stable attitude data (11) together with the accelerometer data (3) to recalibrate the accelerometer data (3) and generate the calibrated accelerometer data (9).

4. said processing unit (8) processes said stabilized attitude data (10); a φ angle (19) of the attitude of the three-axis accelerometer (2) relative to the vehicle's frame of reference, the φ angle (19) representing a first rotation angle about the Z axis of the frame of reference; a θ angle (20) of an attitude of the three-axis accelerometer (2) relative to a reference frame of the vehicle, the θ angle (20) representing a tilt of the reference frame about a Y axis; the processing unit (8) being adapted to compare the change in angle data (21) with a second predetermined threshold (22); - if the change (21) in the angle data is within the second predetermined threshold (22), the processing unit (8) is adapted to divide the angle data (18) into smaller subsection periods and to calculate a moving average (23) of the angle data; or - said processing unit (8) is adapted to generate a median value (24) of said angle data if said variation of said angle data (21) is within said second predetermined threshold (22); 4. The system (1) of claim 3, wherein a processing unit (8) is adapted to process the moving average (23) of the angle data or the median (24) of the angle data together with the accelerometer data (2) to recalibrate the accelerometer data (2) and generate intermediate corrected accelerometer data (9-1) in a horizontal (XY) plane of the vehicle.

5. The processing unit (8) is adapted to compare the intermediate corrected accelerometer data (9-1) with a third predetermined threshold (25) and to compare the orientation data (5) for a corresponding time period with a fourth predetermined threshold (26), and if the intermediate corrected accelerometer (9-1) is above the third predetermined threshold (25) and the orientation data (5) is below the fourth predetermined threshold (26), the processing unit (8) compares a ψ angle (27) of the attitude of the three-axis accelerometer (2) with respect to the reference frame of the vehicle for the stable time period, the ψ angle (27) of the attitude of the three-axis accelerometer (2) with respect to the reference frame of the vehicle for the stable time period with respect to the fourth predetermined threshold (26).

5. The system (1) of claim 4, wherein the processing unit (8) is adapted to determine a ψ angle (27) representing a second rotation about a Z-axis of the quasi-frame, and the processing unit (8) is adapted to determine a most frequently occurring or most likely ψ angle (28) and process the angle data (18) together with the intermediate corrected accelerometer data (9-1) to recalibrate the intermediate corrected accelerometer data (9-1) so that three axes of the acceleration sensor are aligned with the orientation of the vehicle, and generate the final calibrated accelerometer data.

6. 2. The system (1) of claim 1, wherein the processing unit (8) is adapted to compare the corrected accelerometer data (9) in a horizontal (XY) plane of the vehicle with a fifth predetermined threshold (29) at each time point, and if the magnitude of the corrected accelerometer data (9) exceeds the fifth predetermined threshold (29), the processing unit (8) is adapted to determine such time point as an inappropriate driving behavior event (30).

7. 7. The system of claim 6, wherein the processing unit (8) is adapted to determine a first pattern (31) in the corrected accelerometer data (9) in a direction of vehicle movement for consecutive time intervals of different inappropriate driving behavior events (30) and determine the time intervals as intervals of rapid acceleration or deceleration (33) if the first pattern (31) matches a first predetermined pattern (32), and the processing unit (8) is adapted to determine a second pattern (34) in the corrected accelerometer data (9) in a direction perpendicular to the vehicle movement direction for consecutive time intervals of different inappropriate driving behavior events (30) and determine the time intervals as intervals of rapid cornering or rapid lane change (36) if the second pattern (34) exceeds a second predetermined pattern (35).

8. 8. The system (1) of claim 7, wherein the processing unit (8) is adapted to process sections (33) of rapid acceleration or deceleration, and sections (36) of rapid cornering or rapid lane changes, and wherein if sections (33, 36) of the same type are within a predetermined period (37), the processing unit (8) is adapted to combine the sections of the same type and generate a driving problem warning (38) of that type for the combined sections.

9. 9. The system (1) of claim 8, wherein the processing unit (8) is adapted to receive either speed change information (39) or direction change information (40), or a combination thereof, for various inappropriate driving behavior events (30), and to correlate the driving misconduct warning (38) with the speed change information (39) and / or direction change information (40) to determine the accuracy and severity (41) of the driving misconduct warning (38).

10. 7. The system of claim 6, wherein the processing unit (8) is adapted to process the corrected accelerometer data (9) and the direction data (5) for various inappropriate driving behavior events (30), determine a speed (42) of the vehicle during those events, and compare the speed (42) of the vehicle with a first predetermined speed threshold (43), and wherein if the speed (42) of the vehicle exceeds a first predetermined threshold speed (43), the predetermined threshold may be dynamically determined based on a position of the vehicle, the processing unit (8) is adapted to determine the event as an over-speeding event (44).

11. 11. The system of claim 10, wherein the processing unit is adapted to process posture data during periods of unstable posture data to determine a posture data pattern, compare the data pattern to a predetermined posture pattern, and, in case of a match, determine the time interval as a time interval of phone use, and correlate the time interval of phone use with a speed of the vehicle above a second predetermined speed threshold to determine the time interval of phone use as a time interval of distracted driving.

12. 11. The system (1) of claim 10, wherein the processing unit (8) is adapted to process the corrected accelerometer data (9) in the ground direction for short intervals of a period (10) of stable attitude data, determine a data pattern (49) in the corrected accelerometer data, compare the data pattern (49) to a fifth predetermined data pattern (50) and, in case of a match, determine the interval as an interval of bump detection (51), and correlate the interval of bump detection (51) with the speed (42) of the vehicle above a third predetermined speed threshold (43-2) to determine the interval of bump detection as a dangerous crossing event (52) of the bump.

13. 11. The system (1) of claim 10, wherein the processing unit (8) is adapted to process the speed (42) of the vehicle over a short interval to determine sudden speed changes (53), process the corrected accelerometer data (9) around the same time interval to determine an acceleration pattern (54), and correlate the sudden changes in speed (53) and the acceleration pattern (54) to match the acceleration pattern (54) with a sixth predetermined pattern (61) to determine a crash detection event (55).

14. 2. The system (1) of claim 1, wherein the vehicle is adapted to lean sideways, the processing unit (8) is adapted to compare the corrected accelerometer data (9) in the three-dimensional frame of the vehicle at each time point with a ninth predetermined threshold (56), and when the magnitude of the corrected accelerometer data (9) exceeds the ninth predetermined threshold (56), the processing unit (8) is adapted to determine such time point as an inappropriate driving behavior event (30).

15. 15. The system (1) of claim 14, wherein the processing unit (8) is adapted to determine a third pattern (57) in the corrected accelerometer data (9) in a direction of vehicle movement for successive time intervals and to determine the time interval as a section of rapid acceleration or deceleration (33) if the third pattern (57) matches a third predetermined pattern (58), and the processing unit (8) is adapted to determine a fourth pattern (59) in the corrected accelerometer data (9) in the plane perpendicular to the direction of vehicle movement for successive time intervals and to determine the time interval as a section of rapid cornering or abrupt lane change (36) if the fourth pattern (59) exceeds a fourth predetermined pattern (60).

Images