Granted Patent B2

US 12,736,981 · App. 19/101,024 · Granted Sep 15, 2026

Inventors: Tianxiang Su (Cambridge, MA); Suraj Kiran Raman (Cambridge, MA); Muhannad Abdelaziz Abuhaikal (Cambridge, MA); Andriy Gelman (Cambridge, MA); Farshid Mostowfi (Cambridge, MA)

G05D1/644G05D1/247G05D2101/15G05D2105/05G05D2107/73

Abstract

Claims (42)

1 . A downhole robot for intervention, evaluation, or stimulation, comprising:

a driving mechanism;

a power supply operatively connected to the driving mechanism;

a plurality of sensors operatively connected to the power supply;

a trained model for autonomous operation of the downhole robot, the trained model being generated by training the downhole robot to perform operations in a downhole environment, the training comprising generating the trained model using a simulated downhole environment, and the training being performed prior to deployment of the downhole robot;

a compute engine disposed within the downhole robot and operatively connected to the power supply, the compute engine being configured to:

handle processing and computations on-board the downhole robot;

receive at least one wellbore characteristic from the plurality of sensors;

use the trained model for autonomous operation of the downhole robot;

plan one or more actions of the downhole robot based on the trained model; and

cause autonomous execution of the one or more planned actions by the downhole robot in the downhole environment; and

at least one interface configured to transfer data to and from the compute engine.

2 . The downhole robot of claim 1 , wherein the compute engine is a Linux-based system.

3 . The downhole robot of claim 1 , wherein the driving mechanism is a tractor.

4 . A method, comprising:

training an autonomous vehicle to perform operations in a downhole environment, the autonomous vehicle comprising at least one sensor, the training being performed prior to deployment of the autonomous vehicle, the autonomous vehicle comprising a downhole robot, the training comprising:

generating a trained model for autonomous operation of the autonomous vehicle; and

using a simulated downhole environment to generate the trained model;

inserting the autonomous vehicle into the downhole environment;

obtaining at least one wellbore characteristic using data from the at least one sensor in the autonomous vehicle;

processing the at least one wellbore characteristic using a compute engine using the trained model, the compute engine being disposed within the autonomous vehicle;

planning, by the compute engine, one or more actions of the autonomous vehicle based on the trained model; and

causing, by the compute engine, autonomous execution of the one or more planned actions by the autonomous vehicle in the downhole environment.

5 . The method of claim 4 , wherein the at least one wellbore characteristic is one or more of: a distance traveled, a wellbore pressure, or a wellbore temperature.

6 . The method of claim 4 , wherein the autonomous vehicle comprises a tractor configured to move the autonomous vehicle.

7 . The method of claim 4 , wherein the compute engine is a Linux-based system.

8 . The method of claim 4 , wherein the compute engine uses artificial intelligence to control autonomous functions of the autonomous vehicle.

9 . The method of claim 4 , wherein the autonomous vehicle comprises a telemetry module configured to determine positioning of the autonomous vehicle.

10 . The method of claim 4 , further comprising:

performing a calculation to plan the one or more actions by the autonomous vehicle;

determining a required power for the autonomous vehicle to perform the one or more actions by the autonomous vehicle; and

checking a power remaining aboard the autonomous vehicle.

11 . The method of claim 10 , further comprising performing, by the autonomous vehicle, the one or more actions when the power remaining aboard the autonomous vehicle is greater than or equal to the required power for the autonomous vehicle to perform the action one or more actions.

12 . The method of claim 10 , further comprising refusing, by the autonomous vehicle, to perform the one or more actions when the power remaining aboard the autonomous vehicle is less than the required power for the autonomous vehicle to perform the one or more actions.

13 . The method of claim 4 , further comprising saving the at least one wellbore characteristic to a memory of the autonomous vehicle.

14 . The method of claim 4 , wherein the processing the at least one wellbore characteristic using the compute engine using the trained model comprises:

processing acquired logs to infer at least one of an absolute or a relative depth using Bayesian filtering; and

fusing multiple measurements using a sensor fusion-based approach.

15 . The method of claim 4 , wherein the compute engine uses an artificial intelligence planning methodology to plan the one or more actions of the autonomous vehicle.

16 . The method of claim 4 , wherein the compute engine uses constraint satisfaction programming to plan the one or more actions of the autonomous vehicle.

17 . The method of claim 4 , wherein the compute engine uses reinforcement learning to plan the one or more actions of the autonomous vehicle.

18 . The method of claim 4 , wherein the at least one wellbore characteristic processed using the compute engine comprises at least one of: Casing Collar Locator (CCL) Logs, Gamma Ray Logs, Odometry, Pressure and Density Measurements, Acoustic Localization, or Mechanical CCL.

Assignments (1)

Continuity (2)

References Cited (65)