GPS and similar navigation systems rely on on-orbit satellites to triangulate the user's position. Since the distance between the satellite and the mobile user on the ground is too long, this process is inherently susceptible to errors. But Apple believes that it can improve positioning accuracy by applying machine learning to the Kalman estimation filter just released patent application reveal.
The basic concept is that while navigation systems often rely on real-time location determination pings from multiple satellites (which can take precious time during which users may move), machine learning models can be trained to provide users with temporary location estimates Based on data previously collected from the environment. For example, a given city block may have fairly constant satellite signal reflection characteristics, which often introduces errors in user position readings, so machine learning can offset this error.
Although GPS is the most well-known satellite positioning system, Apple's scope of application is not limited to this. It also includes various types of global navigation satellite systems (GNSS), and assumes that in each case, triangulation of raw satellite data will be processed The divided and machine-learned version is transferred to the Kalman linear quadratic estimation filter. The result will be an estimate of the individual locations of the device anytime, anywhere, consistent with data from current user and machine training.
Users using virtual reality and augmented reality headsets may think that the basic premise here is the paradox of machine learning techniques used to correct ground level SLAM tracking In headphones such as Oculus Quest. Although sensors and cameras do a lot of heavy work, Oculus Insight Also used Machine learning model Trained in the homes and environments of multiple early testers to compensate for common local positioning errors and make the wearer's experience smoother.
Many companies have been working to provide more accurate location services and in some cases plan to use 5G cellular network enables accurate mapping at the centimeter level, And introduction double- Either Tri-frequency reception, can receive multiple types of GNSS satellites at the same time. Using machine learning to let the device account for known interference factors in a specific area may add another arrow to the trembling of the navigation system.
Apple's patent application was officially released today after it was quietly filed in August 2019. As a software-based implementation, it can be added to the company's navigation software at any time, or left in the lab indefinitely. Company has Previously acquired companies such as Coherent To increase the speed and accuracy of location services on their devices.