Apple acquired Power by Proxi in October 2017. One of the first patent applications surfacing from Power by Proxi under Apple's name showed that their expertise was with an "object detection system." Such a system is crucial on a multi-device charging pad like AirPower. A user may pull their iPhone out of their pocket to charge it on AirPower and accidentally place along with it an aluminum gum wrapper, some loose change or a transit pass with a magnetic strip on the mat which could cause havoc if not a fire if the charging coils are accidentally overheated. Apple has both acquired and filed patents (01 & 02) working on this problem of foreign objects dumped on a charging pad.

Yesterday a new invention covering this issue surfaced at the US Patent & Trademark Office that reveals that Apple now has a machine-learning-based foreign object detection system.

Apple introduces Machine-Learning-based Foreign Object Detection

Apple's invention relates to advancing their AirPower multi-device charging mat with a superior method of preventing the undesirable heating of foreign metallic objects, like keys or coins, by utilizing a machine-learning-based foreign object detection system.

For example, control circuitry may use inductance measurements and other measurements from the coils in the coil array to determine a probability value indicative of whether a foreign object is present on the charging surface. The control circuitry may use a machine learning classifier to determine the probability of an object on the wireless power transmitting device being a foreign object.

The control circuitry may compare the probability value indicative of whether a foreign object is present on the charging surface to a threshold. In response to determining that the probability value is greater than the threshold, wireless power transmission operations can be blocked to prevent undesirable heating of the foreign objects or other suitable action can be taken.

A user of system 8 (AirPower) shown below may sometimes place one or more wireless power receiving devices on it. For example, a user may place power receiving devices #24A and #24B (an iPhone, Apple Watch and one or more additional devices) on AirPower. Foreign objects #72A, #72B, and #72C which represents items like coins, keys, paper clips, scraps of metal foil, and/or other foreign metallic objects may also be present on AirPower's surface #70.

If coils (#42) are used to transmit wireless power signals while foreign objects are present, eddy currents may be induced in the foreign objects. These eddy currents have the potential to undesirably heat the foreign objects. The foreign objects may include sensitive electronic equipment that could be potentially damaged upon exposure to fields from coils.

As shown in FIG. 3, foreign object 72B is an example of a foreign object that can be detected using image-processing-based foreign object detection. In image-processing based foreign object detection, inductance from each coil in the array may be examined. In-band communication may be used for wireless power transmitting AirPower to receive a device-identifier from each wireless power receiving device on the charging surface.

For example, wireless power receiving device 24A may send a device-identifier identifying device 24A as an iPhone whereas wireless power receiving device 24B may send a device-identifier identifying device 24B as an Apple Watch.  

Each wireless power receiving device may have a characteristic pattern of inductance measurements (and/or quality factors and coupling factors) when the device is present on the charging surface.

For example, the iPhone (#24A) may have a different characteristic pattern of inductance measurements than the Apple watch (#24B). Upon receiving a device identifier, AirPower may compare the known characteristic pattern of inductance measurements of the wireless power receiving device to the present inductance measurements from the coils. If inductance measurements (i.e., inductance measurements corresponding to foreign object 72B) are present that do not correspond to a characteristic pattern from one of the known wireless power receiving devices present, it may be interpreted that a foreign object is present on the charging surface.

Machine-learning-based foreign object detection (sometimes referred to as near-field foreign object detection) may be used to detect foreign objects that are in close proximity to wireless power receiving devices on the charging surface.

Machine-learning-based foreign object detection may include using a machine learning classifier to determine a probability value indicative of whether a foreign object is present on the charging surface. The probability value may be determined using inductance measurements, quality factors, coupling factors, and other desired measurements from the coils. Machine-learning-based foreign object detection may be used to detect foreign objects such as foreign objects 72A and 72C that are in close proximity to a wireless power receiving device on the charging surface.

Foreign Objects on AirPower will trigger various Alerts

In some embodiments, when a foreign object is determined to likely be present, control circuitry may generate an alert that notifies a user that the foreign object is present. The alert may be, for example, a visual alert displayed on an iPhone charging on AirPower or an auditory alert emitted the iPhone.

For example, AirPower may convey the alert to an iPhone on the pad using in-band communication. The iPhone may then display a visual alert using its display, emit an auditory alert using its speakers, or convey a haptic alert using its haptic output device (e.g., a vibrator).

The patent also described a device used by Apple engineers that conducted more than 100,000 trials. In each trial, data such as inductance measurements, quality factors, and coupling factors from the coils were sent to a host device. The host associated data from the coils of AirPower with a desired output of the machine learning classifier (i.e., foreign object detected or no foreign object detected) to train the machine learning classifier.

Apple's patent application that was published today by the U.S. Patent Office was originally filed back in Q1 2018. Considering that this is a patent application, the timing of such a product to market is unknown at this time.

About Making Comments on our Site: Patently Apple reserves the right to post, dismiss or edit any comments. Those using abusive language or negative behavior will result in being blacklisted on Disqus.