Intra-Vehicle Location Finding

Introduction

Southwest Research Institute (SwRI) has found that the accelerometer sensor embedded in most smartphones can determine if the phone is located in a moving vehicle. Smartphone apps can disable distracting functionality when vehicular motion is detected. However, such an app would disable the smartphones of everyone in the car unless a way could be found to distinguish between driver and passenger.

An intra-vehicle location finding system embedded within a phone could locate the phone within the vehicle and thus determine whether the phone was being used by the driver or passenger(s). SwRI collected, analyzed, and modeled driving data to investigate whether typical phone sensors would be sensitive enough to accomplish this task.

Objectives

  • Investigate the accelerations and electromagnetic fields generated by vehicles that may be associated with a position within the vehicle.
  • Determine if the sensors found in smartphones are capable of measuring these accelerations and fields.
  • Discover a series of processing steps and data mining techniques to create models capable of classifying sensor data as “driver” or “passenger.”
image of different views of a car

Technical Approach

One hypothesis for this research was that the acceleration felt by passengers in a vehicle depends on their location within the vehicle. To test this hypothesis the team designed experiments that would measure the accelerations within the vehicle under differing driving scenarios.

To streamline the data collection process the authors developed a custom data collection rig. The rig was composed of four identical sensor nodes connected to a laptop running a custom-built Java data collection application.

The sensors were designed to mimic the sensing capabilities of a smartphone, namely the accelerometer, magnetometer, and GPS. Each node consisted of three main components: an Arduino prototyping board, a 3-axis accelerometer, and a 3-axis magnetometer. A single GPS receiver was used to collect the approximate geolocation for all four nodes. The Arduino board had a microcontroller running custom code that the team wrote to interface with the three sensors. Sensor data was fused by the microcontroller and sent via USB to the base station laptop at regular intervals. A sample interval of 22 ms was chosen (45 Hz update rate).

image of data collection scenarios  
scenarios of turning and braking  
image of graph

Smoothed 3-axis accelerometer data collected during a braking maneuver


image of graph

z-axis accelerometer data collected while traversing a bump


image of a graph

x-axis and z-axis accelerometer data collected when a vehicle is being rocked from side to side


image of a graph

Time domain and frequency domain data collected from a magnetometer

Experiments

Repeatable data collection scenarios were developed so representative data could be recorded for multiple vehicle types. These scenarios contained left turns, right turns, acceleration, and braking. The sensor nodes were placed under the driver and front passenger seats and on both rear passenger seats. Data analysis focused on the driver/passenger sensors for turning scenarios and front/back passenger sensors for the acceleration and braking scenarios.

Using an SwRI test track, data was collected for four different vehicles: a pickup truck, an SUV, and two sedans. Ten instances of each driving scenario were recorded. This data collection effort allowed the team to compare data from different locations in vehicle interiors for different maneuvers across different vehicle types.

Results

Braking can be identified by monitoring the y-axis acceleration data. A braking maneuver begins with a gradual deceleration followed by a more abrupt move in the positive y-direction as the car comes to a stop and oscillates slightly. As the car comes to a stop, the front portion of the car normally moves downward as the back moves upward.

This movement was confirmed for the front and back z-axis accelerometer data. At the moment of stoppage, acceleration in the z-axis is negative for the front sensor and positive for the back sensor.

Driving over a bump in the road can be identified as two positive deviations in the z-axis signal as the front and rear wheels go over the bump. Both front and rear sensors register both bumps; however, the sensor in the front shows a larger deviation for the first bump while the back sensor shows a larger deviation for the second bump.

Another scenario provided evidence that acceleration along the x- and z-axes was either in phase or 180° out of phase depending on whether the sensor was in the driver or passenger position. When a vehicle was rocked around an axis of rotation that was aligned with the vehicle’s y-axis, x- and z-sensor accelerations were in phase if the sensor was on the driver side and 180° out of phase if the sensor was on the passenger side.

Magnetometer data revealed an interesting electromagnetic characteristic in one of the test vehicles, a European-made sedan. The battery for this car was located in the trunk and some cabling ran along the inside of the frame on the lower right passenger side. When the sensor was placed near that side of the vehicle a strong periodic oscillation was observed.

Combining a front vs. back detection method with a passenger-side vs. driver-side detection method enables a determination of whether the sensor is in the driver’s seat or in one of the passenger seats.

Conclusion

All three objectives of this research were accomplished. We determined that:

  • Different positions within a vehicle generate characteristically different sensor data.
  • Inexpensive sensors are fast enough and sensitive enough to detect changes that are usable for intra-vehicle location determination.
  • The modeling techniques developed can locate a sensor in one of the four quadrants of a vehicle.

The next step for this research is to apply the basic concept of intra-vehicle location finding to vehicles other than automobiles, such as boats, tractor trailers, and trains.


Benefiting government, industry and the public through innovative science and technology
Southwest Research Institute® (SwRI®), headquartered in San Antonio, Texas, is a multidisciplinary, independent, nonprofit, applied engineering and physical sciences research and development organization with 10 technical divisions.
07/13/16