Garrett Motion has been granted a patent for a system and method that stores factors in a quadratic programming solver of an embedded model predictive control platform. The system includes a factorization module with a memory containing saved factors, which can be searched to find the nearest stored factor for factor updates. The module also provides variable ordering to reduce the number of factors that need to be stored, allowing for zero floating point operations per unit of time. The method involves precomputing and saving a first set of factors, deducing second factors from the first set, and enabling deduction of third factors with minimal floating-point operations. GlobalData’s report on Garrett Motion gives a 360-degree view of the company including its patenting strategy. Buy the report here.

According to GlobalData’s company profile on Garrett Motion, photovoltaic drones was a key innovation area identified from patents. Garrett Motion's grant share as of June 2023 was 1%. Grant share is based on the ratio of number of grants to total number of patents.

Storing factors in a quadratic programming solver with reduced memory

Source: United States Patent and Trademark Office (USPTO). Credit: Garrett Motion Inc

A recently granted patent (Publication Number: US11687047B2) describes a method for storing factors in a matrix. The method involves precomputing a first set of factors for a matrix that corresponds to a specific set of combinations of currently active constraints. This first set of factors is then saved for future use. The method also includes deducing second factors from the first set of factors using a factor update scheme. Both the first set of factors and the second factors are stored with variable ordering, which is achieved using a permutation matrix.

One of the key advantages of this method is that it enables the deduction of third factors with zero floating-point operations per unit time (FLOPs) by simply removing a row and/or column. This reduces the computational complexity and improves efficiency. Additionally, the method utilizes a factor update with a non-zero cost in FLOPs to further reduce memory consumption.

The patent also mentions that the method can be used in embedded model predictive control, which is a technique used in various applications such as robotics and process control. Furthermore, the method can be implemented using a Newton step/gradient projection approach, which is a common method in optimization algorithms.

Overall, this patented method provides an efficient way to store factors in a matrix by precomputing and deducing factors, utilizing variable ordering, and reducing computational complexity and memory consumption. Its applications in embedded model predictive control and use of established optimization techniques make it a valuable innovation in the field.

To know more about GlobalData’s detailed insights on Garrett Motion, buy the report here.