Tesla patents method to simplify math for robot brain chips
Tesla describes a way to convert complex decimal numbers into simpler, smaller formats while a robot is running. This process, called quantization, helps the robot's computer process data faster by using less memory and power.
In the filing’s own words
Disclosed herein are systems and methods for quantizing floating-point values during operation of robotic systems. For example, a system can be configured to receive input data having multiple values represented in a floating-point format and determine a maximum absolute value from these values. The system can adjust the representation of each value by removing at least one bit from the exponent portion and redistributing it to the mantissa portion. In examples, the system can also extract a sign bit and align each value by shifting each bit. Based on the adjusted values, the system can perform dynamic quantization of the input data based at least in part on a scaling value determined using the maximum absolute value. The input data can then be provided as an input to at least a portion of a neural network during execution of one or more operations by the neural network.
What exactly they claim (claim 1)
1. A system for quantizing floating-point values during operation of an automated robotic system, the system comprising: one or more processors configured to: obtain input data comprising a first set of values represented using a first data format during operation of an automated robotic system; determine a maximum absolute value from among the first set of values based on the input data; in response to determining the maximum absolute value, adjust each value of the first set of values based on the maximum absolute value; quantize the first set of values based on an exponent and a mantissa of each value to generate a quantized set of values having a second data format; provide at least a portion of the quantized set of values to a neural network to cause the neural network to generate an output based on the quantized set of values; and control operation of the automated robotic system based on the output of the neural network.
Facts
- Company
- Tesla (TESLA, INC.)
- Official title
- SYSTEMS AND METHODS FOR QUANTIZING FLOATING-POINT VALUES DURING OPERATION OF AN AUTOMATED ROBOTIC SYSTEM
- Publication number
- US20260299533A1
- Application number
- 19629949
- Filed
- Mar 26, 2026
- Published
- Oct 1, 2026
- Inventors
- 4
- Claims
- 20
Classified under
G05B 13/027 Physics & computingB60W 50/00 Operations & transportB60W 2050/0028 Operations & transportB60W 60/00 Operations & transportSummaries are simplified from the filing’s own abstract and first claim. A patent filing is not a product announcement.
Patent data and drawings: U.S. Patent and Trademark Office (USPTO). Nebula is not affiliated with or endorsed by the USPTO.