# US 2018/0026573 A1 · Machine learning device which learns current command for motor, motor controller, and machine learning method > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 2018/0026573 A1 - Document ID: US20180026573A1 - Status: Published application - Kind: A1 - Application number: 15/628,641 - Filed: 2017-06-21 - Published: 2018-01-25 - Earliest claimed: 2016-07-21 - Assignees: Fanuc Corporation - Applicants: Fanuc Corporation - Inventors: Koudai AKASHI; Hidetoshi UEMATSU - Web page: https://patentgrep.com/patent/US20180026573A1 ## CPC - H02P 29/60: Controlling or determining the temperature of the motor or of the drive (H02P29/02 takes precedence) - G06N 3/006: based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO] - G06N 3/02: Neural networks - G06N 3/084: Backpropagation, e.g. using gradient descent - H02P 21/0025: {implementing a off line learning phase to determine and store useful data for on-line control} - H02P 21/13: Observer control, e.g. using Luenberger observers or Kalman filters - H02P 29/10: for preventing overspeed or under speed - H02P: CONTROL OR REGULATION OF ELECTRIC MOTORS, ELECTRIC GENERATORS OR DYNAMO-ELECTRIC CONVERTERS; CONTROLLING TRANSFORMERS, REACTORS OR CHOKE COILS - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract A machine learning device which learns a current command for a motor, the machine learning device including a state observation unit which observes a state variable including a motor rotation speed or a motor torque command of the motor and at least one of a motor torque, a motor current, and a motor voltage of the motor; and a learning unit which learns the current command for the motor based on the state variable. ## Claims ### Claim 1 A machine learning device which learns a current command for a motor, the machine learning device comprising: a state observation unit which observes a state variable including a motor rotation speed or a motor torque command of the motor and at least one of a motor torque, a motor current, and a motor voltage of the motor; and a learning unit which learns the current command for the motor based on the state variable observed by the state observation unit. ### Claim 2 (depends on 1) The machine learning device according to claim 1, further comprising a decision unit which determines the current command for the motor based on an output of the learning unit. ### Claim 3 (depends on 1) The machine learning device according to claim 1, wherein the motor is a three-phase alternating-current synchronous motor, and the current command for the three-phase alternating-current synchronous motor includes a d-axis current command and a q-axis current command which are electrically and magnetically orthogonal to each other. ### Claim 4 (depends on 3) The machine learning device according to claim 3, wherein learning of the d-axis current command and the q-axis current command is performed in terms of a rotor phase angle of the three-phase alternating-current synchronous motor. ### Claim 5 (depends on 1) The machine learning device according to claim 1, wherein the learning unit includes: a reward calculation unit which calculates a reward based on an output of the state observation unit; and a value function update unit which updates a value function relating to the current command for the motor based on the output of the state observation unit and an output of the reward calculation unit. ### Claim 6 (depends on 5) The machine learning device according to claim 5, wherein for each motor rotation speed or motor torque command for the motor, the reward calculation unit increases the reward when the motor torque increases, decreases the reward when the motor current exceeds a predetermined value, and decreases the reward when the motor voltage is saturated. ### Claim 7 (depends on 1) The machine learning device according to claim 1, wherein the learning unit includes: an error calculation unit which calculates an error based on the output of the state observation unit and supervised data; and an error model update unit which updates an error model relating to the current command for the motor based on the output of the state observation unit and an output of the error calculation unit. ### Claim 8 (depends on 1) The machine learning device according to claim 1, wherein the machine learning device further comprises a neural network. ### Claim 9 (depends on 1) The machine learning device according to claim 1, wherein the current command for the motor as learned by the machine learning device is configured in such a manner that data is mutually shared or exchanged among the plurality of machine learning devices via a communication medium. ### Claim 10 (depends on 1) The machine learning device according to claim 1, wherein the machine learning device is located on a cloud server. ### Claim 11 (depends on 1) A motor controller comprising: the machine learning device according to claim 1; a motor drive unit which drives and controls the motor; and a motor control unit which controls the current command for the motor through the motor drive unit based on the current command for the motor as learned by the machine learning device. ### Claim 12 (depends on 11) The motor controller according to claim 11, wherein the state observation unit observes a state variable including the motor rotation speed or the motor torque command of the motor, the motor current, and the motor voltage, and the motor torque is calculated from the motor current and the motor voltage as observed. ### Claim 13 A machine learning method of learning a current command for a motor, the machine learning method comprising: observing a state variable including a motor rotation speed or a motor torque command of the motor and at least one of a motor torque, a motor current, and a motor voltage for the motor; and learning the current command for the motor based on the state variable as observed. ### Claim 14 (depends on 13) The machine learning method according to claim 13, wherein learning of the current command for the motor includes: calculating a reward based on the state variable as observed; and updating a value function relating to the current command for the motor based on the state variable as observed and the reward as observed. ### Claim 15 (depends on 14) The machine learning method according to claim 14, wherein in the reward calculation, for each motor rotation speed or motor torque command, the reward is increased when the motor torque increases, the reward is decreased when the motor current exceeds a predetermined value, and the reward is decreased when the motor voltage is saturated.