Evolutionary control of machine based on user's preference inferred from user's operation

• US 6,314,412 B1
• Filed: 09/24/1998
• Issued: 11/06/2001
• Est. Priority Date: 09/29/1997
• Status: Expired due to Fees

• Current Assignee: Yamaha Hatsudoki Kabushiki Kaisha (Yamaha Corporation)
• Original Assignee: Yamaha Hatsudoki Kabushiki Kaisha (Yamaha Corporation)
• Inventors: Yamaguchi, Masashi, Kamihira, Ichikai
• Primary Examiner(s): Chaki, Kakali
• Assistant Examiner(s): Khatri, Anil
• Application Number:
  US09/159,836
• Time in Patent Office: 1,139 Days
• Field of Search: 706/12, 706/13, 706/14, 706/23, 706/25
• US Class Current: 706/13
• CPC Class Codes: G05B 13/0265 the criterion being a learning criterion
• CPC Class Codes: G05B 13/027 using neural networks only
• CPC Class Codes: G06N 3/086 using evolutionary algorithms, e.g. genetic algorithms or genetic programming

Abstract

An evolutionary control for a subject such as an engine installed in a vehicle is conducted by the steps of: selecting coefficients, as genes, affecting the control characteristics of the control system; creating plural control units as chromosomes, each being constituted by plural genes; expressing the genes of each chromosome as control characteristics by controlling the subject using the control system having the control characteristics; selecting at least one chromosome from the created chromosomes based on the control characteristics expressed by the genes in view of the user'"'"'s preference; and causing the at least one chromosome to evolve using a genetic algorithm, thereby obtaining control characteristics suitable to the user. In this method, the characteristics of the product can effectively be adjusted after its purchase based on the user'"'"'s preference.

Claims

1. A method for real-time evolutionary control of an machine operated by a users, adaptive to the user based on the user'"'"'s preference inferred from the user'"'"'s operation of the machine, said machine being controlled by a control system, said control system outputting control signals for controlling the machine when receiving input signals, the relationship between the control signals and the input signals being defined by coefficients, said method comprising the steps of:

(a) selecting plural sets of values of coefficients to create plural potential control units in the control system;

(b) controlling the machine by using the control system with each potential control unit to evaluate control characteristics of each potential control unit;

(c) inferring the user'"'"'s preference for control characteristics from changes in the user'"'"'s operation of the machine;

(d) selecting at least one potential control unit based on the user'"'"'s preference for its control characteristics;

(e) subjecting said at least one potential control unit to evolutionary computing to create plural offspring potential control units, wherein each set of coefficient values are used as genes, each potential control unit is created as a chromosome, control characteristics expressed by each chromosome constitute an individual, and selection of individuals is conducted based on the user'"'"'s preference; and

(f) repeating steps (b) to (e) until the control system obtains control characteristics adapted to the user.

2. The method according to claim 1, wherein the relationship between the control signals and the input signals of the control system is constituted by a neural network having input terminals for the input signals and output terminals for the control signals, and wherein the coefficients are coupling coefficients of the network.

3. The method according to claim 1, further comprising saving the selected potential control unit as a fit control unit in a memory, and using the fit control unit.

4. The method according to claim 3, further comprising the steps of:

(g) while using the adapted control unit stored in the memory to control the machine, repeating steps (a) through (d) to select a second adapted control unit;

(h) saving the second adapted control unit in a second memory;

(i) switching from the first adapted control unit to the second adapted control unit to be used;

(j) using the second adapted control unit stored in the second memory to control the subject;

(k) repeating steps (b) through (f) to select a third adapted control unit and saving the same in the first memory; and

(l) repeating steps (g) through (k).

5. The method according to claim 1, wherein the user'"'"'s preference is measured by smoothness of the user'"'"'s operation.

6. The method according to claim 5, wherein the smoothness is determined based on a change in degree of the operational action by the user and/or the frequency of the operational action by the user.

7. The method according to claim 1, wherein, in step (a), the plural sets of values of the coefficients are selected by performing steps (a) through (e) using a computer simulation of the machine without operating the machine, wherein the user'"'"'s preference is predetermined in the simulation.

8. The method according to claim 1, wherein the machine is an engine for a vehicle, and the control characteristics of the control system are selected from the group consisting of its fuel-injection quantity, ignition timing, electric-throttle angle, timing of intake and exhaust valves, valve drift, and timing of intake and exhaust control valves.