Omnidirectional wheeled humanoid robot based on a linear predictive position and velocity controller

• US 10,232,508 B2
• Filed: 04/17/2015
• Issued: 03/19/2019
• Est. Priority Date: 04/17/2014
• Status: Active Grant

• Current Assignee: Institut National de Recherche en informatique et en Automatique, SoftBank Robotics Europe SAS
• Original Assignee: Institut National de Recherche en informatique et en Automatique, SoftBank Robotics Europe SAS
• Inventors: Lafaye, Jory, Collette, Cyrille, Wieber, Pierre-Brice
• Primary Examiner(s): Sample, Jonathan L
• Application Number:
  US15/300,221
• Publication Number:
  US 20170144299A1
• Time in Patent Office: 1,432 Days
• Field of Search: 700245-264
• CPC Class Codes: B25J 5/007 mounted on wheels
• CPC Class Codes: B25J 9/162 Mobile manipulator, movable base with manipulator arm mounted on it
• CPC Class Codes: B25J 9/1628 characterised by the control loop
• CPC Class Codes: G05D 1/0891 specially adapted for land vehicles

Abstract

A humanoid robot with a body joined to an omnidirectional mobile ground base, equipped with: a body position sensor, a base position sensor and an angular velocity sensor to provide measures, actuators comprising at least 3 wheels located in the omnidirectional mobile base, extractors for converting sensored measures into useful data, a supervisor to calculate position, velocity and acceleration commands from the useful data, means for converting commands into instructions for the actuators, wherein the supervisor comprises: a no-tilt state controller, a tilt state controller and a landing state controller, each controller comprising means for calculating, position, velocity and acceleration commands based on a double point-mass robot model with tilt motion and on a linear model predictive control law, expressed as a quadratic optimization formulation with a weighted sum of objectives, and a set of predefined linear constraints.

Claims

1. A humanoid robot with a body joined to an omnidirectional mobile ground base, equipped with:

a body position sensor, a base position sensor and an angular velocity sensor to provide measures,

actuators comprising joints motors and at least 3 wheels located in the omnidirectional mobile base, with at least one omnidirectional wheel,

extractors for converting sensored measures,

a supervisor to calculate position, velocity and acceleration commands from the extracted data,

means for converting commands into instructions for the actuators,

wherein the supervisor comprises;

a no-tilt state controller, a tilt state controller and a landing state controller, each controller comprising means for calculating from the extracted data, pre-ordered position and velocity references, and a tilt angle and angular velocity references set to 0, position, velocity and acceleration commands based on a double point-mass robot model and on a linear model predictive control law with a discretized time according to a sampling time period T and a number N of predicted samples, expressed as a quadratic optimization formulation with a weighted sum of objectives, and a set of predefined linear constraints,

an impact angular velocity and a landing impact time estimator and

means for choosing a controller according to an estimated impact angular velocity and extracted angular velocity.

2. The humanoid robot of claim 1, wherein the base has a constant angular acceleration.

3. The humanoid robot of claim 1, wherein the no-tilt controller is based on a no-tilt robot model and is able to calculate position, velocity and acceleration commands from the extracted data using pre-ordered references, and in that the objectives are:

a base position objective,

a base velocity objective,

an objective related to the distance between the CoP and the base center, CoP being the barycenter of contact forces between the robot and the ground,

and the constraints are;

a maximum velocity and acceleration of the mobile base,

a CoP limit.

4. The humanoid robot of claim 1, wherein the tilt controller is based on a robot model with tilt motion and is able to calculate position, velocity and acceleration commands from the extracted data using pre-ordered references and a tilt angle and angular velocity references set to 0, and in that the objectives are tilt angle minimization and angular velocity minimization and the constraints are kinematic limits of the mobile base, kinematic limits of the body, a positive tilt angle and a move of the body only on the angular velocity axis.

5. The humanoid robot of claim 3, wherein a weighted numerical stability objective is added to the weighted sum of objectives.

6. The humanoid robot of claim 1, wherein the landing controller is based on a no-tilt robot model and is able to calculate position, velocity and acceleration commands from the extracted data using pre-ordered references and a tilt angle and angular velocity references set to 0, and in that the objectives are an objective related to the distance between the CoP and the base center, CoP being the barycenter of contact forces between the robot and the ground, and a numerical stability objective, the constraints are a maximum velocity and acceleration of the mobile base and kinematics limits of the body and a CoP limit and a move of the body only on the angular velocity axis.

7. The humanoid robot of claim 1, wherein a base velocity objective is added to the weighted sum of objectives.

8. The humanoid robot of claim 3, wherein the constraints comprise kinematics limits of the body.

9. A method for controlling a humanoid robot with a body joined to an omnidirectional mobile ground base, with actuators comprising at least three wheels with at least one omnidirectional wheel comprising:

retrieving position measure of the body, position measure of the base, tilt angle of the robot and angular velocity measure of the robot, at pre-defined sampling times,

converting these measures in extracted data,

estimating an impact angular velocity and a landing impact time;

choosing a state of the robot among a defined tilt-state, or no-tilt state or landing state of the robot according to said impact angular velocity and landing impact time;

using the extracted data, and, according to said state of the robot pre-ordered position and velocity references, and a title angle and angular velocity references set to 0, calculating position, velocity and acceleration commands based on a double point-mass robot model and on a linear model predictive control law with a discretized time according to a sampling time period T and a number N of predicted samples, and expressed as a quadratic optimization formulation with a weighted sum of objectives with predefined weights and a set of linear constraints using respectively;

a tilt state control law if the state of the robot is the tilt-state;

a no-tilt state control law if the state of the robot is the no-tilt state;

a landing state control law if the state of the robot is the landing state;

converting these commands into instructions for the robot actuators; and

sending said instructions to said robot actuators.

10. The method of claim 9, wherein the state of the robot is defined according to the following steps:

if one among the tilt angle measure and/or the angular velocity measure is greater than zero, estimating an estimated impact angular velocity and an estimated impact time,

initially, placing the robot is in the no-tilt state,

when the robot is in the no-tilt state;

if there is no estimated impact time, switching to the tilt state;

if the impact angular velocity is above a pre-defined limit1, switching to the tilt state;

if the measured angular velocity is above a pre-defined limit2, switching to the tilt state;

else, if the tilt angle is not null, switching to the landing state,

if the tilt angle is null staying in the no-tilt state,

when the robot is in the tilt state;

if the measured angular velocity is above the pre-defined limit2, staying in the tilt state;

if the estimated impact velocity is under a pre-defined limit1, switching to the landing state;

else, staying in tilt state,

when the robot is in the landing state;

if there is no estimated impact time, switching to the tilt state;

if the impact angular velocity is above a pre-defined limit1, switching to the tilt state;

if the measured angular velocity is above a pre-defined limit2, switching to the tilt state;

if the tilt angle falls to 0, and if an angular velocity measured at a next time is null, switching to the non-tilt state, else, staying in the landing state.

11. A computer program linked to actuators of a humanoid robot with a body joined to an omnidirectional mobile ground base, with actuators comprising at least three wheels with at least one omnidirectional wheel, said computer program comprising computer code fit for executing when running on a computer a method comprising:

retrieving position measure of the body, position measure of the base, tile angle of the robot and angular velocity measure of the robot, at pre-defined sampling times,

converting these measures in extracted data,

estimating an impact angular velocity and a landing impact time,

choosing a state of the robot among a defined tilt-state, or non-tilt state or landing state of the robot according to said impact angular velocity and landing impact time,

using the extracted data, and, according to said state of the robot pre-ordered position and velocity references, and a tilt angle and angular velocity references set to 0, calculating position, velocity and acceleration commands based on a double point-mass robot model and on a linear model predictive control law with a discretized time according to a sampling time period T and a number N or predicted samples, and express as a quadratic optimization formulation with a weighted sum of objectives with predefined weights and a set of linear constraints using respectively;

a tilt-state control law if the state of the robot is the tilt state;

a no-tilt state control law if the state of the robots is the no-tilt state;

a landing state control law if the state of the robot is the landing state;

converting these commands into instructions for the robot actuators;

sending said instructions to said robot actuators.