Machine-learning based vehicle diagnostics and maintenance
In general, techniques are described by which provide vehicle diagnostics and maintenance. A device comprising an interface and a processor may be configured to perform the techniques. The interface may be configured to communicate with a plurality of sensors to obtain a plurality of sensor signals representative of one or more states of a vehicle. The processor may be configured to apply a trained classifier with respect to the plurality of sensor signals to identify one or more components of the vehicle that result in a change of the vibration during operation of the vehicle.
1. A device configured to perform diagnostics, the device comprising:
means for obtaining, from a plurality of sensors, a plurality of sensor signals representative of one or more states of a vehicle, wherein the plurality of sensors are not directly coupled to a wheel of the vehicle and none of the plurality of sensor signals defines a tire pressure of the wheel of the vehicle;
means for providing an indirect tire pressure monitoring system (iTPMS) to indirectly identify a low tire pressure state based on the plurality of sensor signals; and
means for applying a trained classifier to the plurality of sensor signals to identify one or more components of the vehicle that result in a change of vibration during operation of the vehicle and whether a cause of the vibrations during the operation of the vehicle is the low tire pressure state of the wheel of the vehicle, wherein the trained classifier includes both a trained random forest classifier trained using supervised learning and a trained hidden Markov model trained using unsupervised learning.
2. A method of performing diagnostics, the method comprising:
obtaining, from a plurality of sensors, a plurality of sensor signals representative of one or more states of a vehicle, wherein the plurality of sensors are not directly coupled to a wheel of the vehicle and none of the plurality of sensor signals defines a tire pressure of the wheel of the vehicle;
providing an indirect tire pressure monitoring system (iTPMS) to indirectly identify a low tire pressure state based on the plurality of sensor signals; and
applying a trained classifier to the plurality of sensor signals to identify one or more components of the vehicle that result in a change of vibration during operation of the vehicle and whether a cause of the vibrations during the operation of the vehicle is the low tire pressure state of the wheel of the vehicle, wherein the trained classifier includes both a trained random forest classifier trained using supervised learning and a trained hidden Markov model trained using unsupervised learning.
3. The method of claim 2 , wherein applying the trained classifier comprises applying the trained classifier with respect to the plurality of sensor signals to also identify one or more components of a chassis of the vehicle that result in the change of the vibration during the operation of the vehicle.
4. The method of claim 3 , wherein the components of the chassis of the vehicle includes one or more of a damper, a spring, a wheel hub, a brake, a control arm, a bushing, a tire rod, and a wheel bolt.
5. The method of claim 2 ,
wherein the one or more sensors include a plurality of different types of sensors configured to at least some of sense wheel speed, relative wheel speed, yaw rate, pitch, lateral acceleration, lateral speed, roll, longitudinal acceleration, longitudinal speed, engine speed, or engine torque, and
wherein the plurality of sensor signals include two or more sensor signals representative of two or more of wheel speed, relative wheel speed, yaw rate, pitch, lateral acceleration, lateral speed, roll, longitudinal acceleration, longitudinal speed, engine speed, and engine torque.
6. The method of claim 2 , wherein applying the trained random forest classifier of the trained classifier comprises:
applying one or more decision trees to the plurality of sensor signals to obtain one or more results;
summing the one or more results; and
identifying the one or more components of the vehicle that result in the change of the vibration during operation of the vehicle based on the summed results.
7. The method of claim 2 , further comprising transforming, prior to applying the trained classifier, one or more of the plurality of sensor signals from a spatial domain to a frequency domain to obtain one or more frequency domain sensor signals,
wherein applying the trained classifier comprises applying the trained classifier to the one or more frequency domain signals to identify the one or more components of the vehicle that result in the change of the vibration during the operation of the vehicle.
8. The method of claim 2 , further comprising:
obtaining, based on the one or more components, a notification configured to inform an occupant of the vehicle of the one or more components of the vehicle that result in the change in the vibration during the operation of the vehicle; and
interfacing with a display to present the notification.
9. A device configured to perform diagnostics, the device comprising:
a hardware interface configured to communicate with a plurality of sensors to obtain a plurality of sensor signals representative of one or more states of a vehicle, wherein the plurality of sensors are not directly coupled to a wheel of the vehicle and none of the plurality of sensor signals defines a tire pressure of the wheel of the vehicle; and
one or more processors configured to:
provide an indirect tire pressure monitoring system (iTPMS) to indirectly identify a low tire pressure state based on the plurality of sensor signals; and
apply a trained classifier with respect to the plurality of sensor signals to identify one or more components of the vehicle that result in a change of vibration during operation of the vehicle and whether a cause of the vibrations during the operation of the vehicle is the low tire pressure state of the wheel of the vehicle, wherein the trained classifier includes both a trained random forest classifier trained using supervised learning and a trained hidden Markov model trained using unsupervised learning.
10. The device of claim 9 , wherein the one or more processors are configured to apply the trained classifier with respect to the plurality of sensor signals to also identify one or more components of a chassis of the vehicle that result in the change of the vibration during the operation of the vehicle.
11. The device of claim 9 ,
wherein the one or more sensors include a plurality of different types of sensors configured to at least some of sense wheel speed, relative wheel speed, yaw rate, pitch, lateral acceleration, lateral speed, roll, longitudinal acceleration, longitudinal speed, engine speed, or engine torque, and
wherein the plurality of sensor signals include two or more sensor signals representative of two or more of wheel speed, relative wheel speed, yaw rate, pitch, lateral acceleration, lateral speed, roll, longitudinal acceleration, longitudinal speed, engine speed, and engine torque.
12. The device of claim 9 , wherein the one or more processors are further configured to transform, prior to applying the trained classifier, one or more of the plurality of sensor signals from a spatial domain to a frequency domain to obtain one or more frequency domain sensor signals, and
wherein the one or more processors are configured to apply the trained classifier to the one or more frequency domain signals to identify the one or more components of the vehicle that result in the change of the vibration during the operation of the vehicle.
13. The device of claim 9 , wherein the one or more processors are further configured to:
obtain, based on the one or more components, a notification configured to inform an occupant of the vehicle of the one or more components of the vehicle that result in the change in the vibration during the operation of the vehicle; and
interface with a display to present the notification.
14. A non-transitory computer-readable medium having stored thereon instructions that, when executed, cause one or more processors of a diagnostic device to:
interface with a plurality of sensors to obtain a plurality of sensor signals representative of one or more states of a vehicle, wherein the plurality of sensors are not directly coupled to a wheel of the vehicle and none of the plurality of sensor signals defines a tire pressure of the wheel of the vehicle;
provide an indirect tire pressure monitoring system (iTPMS) to indirectly identify a low tire pressure state based on the plurality of sensor signals; and
apply a trained classifier with respect to the plurality of sensor signals to identify one or more components that result in a change of vibration during operation of the vehicle and whether a cause of the vibrations during the operation of the vehicle is the low tire pressure state of the wheel of the vehicle, wherein the trained classifier includes both a trained random forest classifier trained using supervised learning and a trained hidden Markov model trained using unsupervised learning.