Image Image Image Image Image Image Image Image Image Image
Scroll to top

Top

No Comments

Patent of the Month – Smartphone Detection Of Vehicle Maneuvers

Patent of the Month – Smartphone Detection Of Vehicle Maneuvers

| On 04, Feb 2018

Darrell Mann

There’s an awful lot of talk these days about autonomous vehicles. A lot of it is naïve. We won’t all be driving around in autonomous vehicles in the next five years – or even ten – because there are too many legal, moral and ethical contradictions to be worked through. The fact that over 3% of the population currently earn a living by driving might also be a factor: where are those millions of people going to go? That said, we know it will happen eventually. Which means that any interim solutions can only ever be just that: interim.

This month’s patent of the month is such a solution. It originates from a pair of inventors at the University of Michigan. US8,834,222 was granted on December 5. Mention the word ‘Michigan’ and we automatically think ‘automotive’. That said, the patent highlights an increasingly familiar story in the automotive industry: the winners are the non-automotive providers. In their bid to work towards autonomous vehicles, the industry has tended to focus internally, and has tried to engineer solutions using resources they can add to a vehicle. The problem is that the best and most appropriate resources already exist in the smartphones we all carry around. Which means that no matter what the automotive industry might try and do, they find themselves fighting an ever more difficult to win battle. So, while the automotive OEMs are busy fitting expensive camera and sensor systems to their top end vehicles, the University of Michigan trump them all by demonstrating that an everyday smartphone can already do what they’re trying to do.

Here’s what the inventors have to say about the problem under consideration:

Automobiles bring a wide range of conveniences as well as fatalities. In 2012, the reported number of fatalities from road accidents was 30,800 in the U.S. alone. Of these fatalities, 23.1% involved lane control–i.e., merging or changing lines or driving on curvy roads–and 7.7% involved turning maneuvers, i.e., turning left/right or making U-turns. In total 30.8% of the fatalities were related to vehicle steering maneuvers.

As most of these fatalities had resulted from the driver’s careless or erroneous steering, those accidents could have been minimized or prevented if effective safety mechanisms had been deployed in the vehicles. There has been an on-going push for incorporating electronic safety features in vehicles to assist drivers’ steering.

Many existing driving assistance systems, such as lane-departure warning or lane-keeping assistance, are exploiting the advanced built-in sensors (e.g., cameras, radars, and infrared sensors), or utilizing existing sensors on smartphones to detect the steering maneuvers in order to assist driving. However, these approaches suffer two limitations: many built-in sensors are only available on newer high-end cars, and hence such safety solutions cannot be applied to a wide range of type/year models of cars; and both built-in sensors and smartphone-based applications overly rely on cameras to get road information. Although such systems claim that smartphone cameras are sufficient in assisting the driver, they have limitations in terms of computational overhead and inaccuracy. The accuracy of camera-based approaches depends on visibility, thus render ineffective in various driving conditions with limited visibility.

Alternatively, smartphone sensors, such as gyroscope, accelerometer, magnetometer, etc., can be exploited to detect vehicle steering maneuvers and thus perform the same steering-assistance functionalities that would have been achieved with use of cameras. These camera-free approaches have advantages of requiring much less computational resources and power, and also being immune to visibility distortions. However, it is known to be very difficult to differentiate the steering maneuvers, which is one of the main reasons for camera-based approaches being most prevalent.

Which looks something like this when mapped on to the Contradiction Matrix:

The Inventive Principle suggestions for which seem to align very elegantly with the solution generated by the inventors. Here’s what they describe:

A method is provided for detecting vehicle maneuvers using a mobile phone [Principle 24 – Intermediary] residing in the vehicle. The method includes: receiving signals from two or more sensors residing in the mobile phone while the vehicle is moving, where the signals are indicative of vehicle movement and reported relative [Principle 37 – Relative Change] to a coordinate system associated with the mobile phone [Principle 26 – Copying]; converting the signals from the two or more sensors from the phone coordinate system to a geographic coordinate system; deriving a detection signal indicative of angular speed of the vehicle about a vertical axis of the geographic coordinate system, where the detection signal is derived from the converted signals; correlating the detection signal to one of a plurality of predefined signatures [Principle 10 – Prior Action], where each signature represents a different maneuver made by a vehicle; determining a direction of travel for the vehicle; determining a horizontal displacement of the vehicle relative to the direction of travel; and identifying a maneuver made by the vehicle based on the correlation of the detected signal to one of the plurality of predefined signatures and magnitude of the horizontal displacement of the vehicle. The geographic coordinate system includes a vertical axis defined in relation to surface the vehicle is traversing upon and the vertical axis does not align with any of the axes that define the phone coordinate system.

Correlating the detection signal to one of a plurality of predefined signatures includes identifying one or two bumps [Principle 18 – Vibration] in the detection signal, where one bump correlates to turn by the vehicle and two bumps correlates to a lane change by the vehicle.

Vehicle maneuvers may be classified as either a lane change, a turn or a curvy road. Identifying a maneuver made by the vehicle includes comparing magnitude of the horizontal displacement to a threshold; and classifying the maneuver as a curvy road in response to a determination that the magnitude of the horizontal displacement exceeds the threshold. In one aspect, horizontal displacement is determined by determining velocity of the vehicle along the direction of travel; determining angular speed of the vehicle about a vertical axis of the geographic coordinate system; and determining horizontal displacement as a function of velocity and angular speed of the vehicle. Additionally, a turn may be further classified as either a right turn, a left turn or a U-turn based on a difference in vehicle heading angle between start and end of vehicle maneuver.

If reading through patents is not your thng, check out a very followable presentation of the solution here: http://slideplayer.com/slide/4881856/.

Meanwhile, sit back and admire disruption in action.

Submit a Comment