Basic RT components and utilities: $title

Overview

This component is generate walking pattern and control Center Of Gravity (COG) for legged robots without sensor feedback.

COG control

This component modifies input joint angles to track reference COG position. This component receives reference joint angles from "qRef" inport. Then it solves Inverse Kinematics (or something) to track reference COG position, obtains jonint angles, and outputs the joint angles as "q" outport.

In this calculation, this component tracks reference XY COG position, reference Z base position, and reference base orientation.

Mode

This component has modes and behaves as Finite State Machine. By default, this component is idle mode ("MODE_IDLE"), in which it do not modify input joint angles. When OpenHRP::AutoBalancerService::startAutoBalancer() are called in IDLE mode, it switches to controlling mode ("MODE_ABC"), in which it modifies joint angles. When OpenHRP::AutoBalancerService::stopAutoBalancer() are called in controlling mode, it switches to idle mode.

Controlled end-effectors

User can specify which end-effector is controlled by the argument of OpenHRP::AutoBalancerService::startAutoBalancer(). When OpenHRP::AutoBalancerService::startAutoBalancer() are called with ["rleg", "lleg"], it controlles reference end-effector position and orientation for "rleg" and "lleg". When OpenHRP::AutoBalancerService::startAutoBalancer() are called with ["rleg", "lleg", "rarm", "larm"], it controlles reference end-effector for "rleg", "lleg", "rarm", "larm". In this case, the robot seems to keep both feet and hands position and orientation.

Walking pattern generation

This component has also the feature as walking pattern generation, named as GaitGenerator. When OpenHRP::AutoBalancerService::goPos(), OpenHRP::AutoBalancerService::goVelocity(), OpenHRP::AutoBalancerService::setFootSteps() are called, this component starts to use reference XY COG position from "GaitGenerator". For OpenHRP::AutoBalancerService::goPos() and OpenHRP::AutoBalancerService::setFootSteps(), this component stops to use it after completing walking command. After OpenHRP::AutoBalancerService::goVelocity() is called, this component continuously generates walking pattern. When OpenHRP::AutoBalancerService::goStop() are called in this case, this completing stops to use it after completing walking command.

Preview Controller Example

Preview controller example, in which trajectories are displayed on graphs (COG, ZMP, ... etc).

testPreviewController # Non-ROS environemnt
rosrun hrpsys testPreviewController # ROS environemnt

GaitGenerator Example

GaitGenerator example, in which trajectories are displayed on graphs (COG, ZMP, foot trajectories ... etc).

Users can change test type (fwd-walking example, rotation-walking example, stair-walking, toe-heel example, ... etc).

Users can also change GaitGenerator parameters (step time, step height, orbit type, ... etc).

testGaitGenerator [test-type-option] [gg-parameters-options] # Non ROS environment
rosrun hrpsys testGaitGenerator [test-type-option] [gg-parameters-options] # ROS environment

[test-type-option] is test type and [gg-parameters-options] are GaitGeneratorParameters. To learn test type, please execute testGaitGenerator without arguments.

implementation_id	AutoBalancer
category	example

About Toe Joints

AutoBalancer can support toe joints. Toe joint is defined as end-link joint in the case that end-effector link != force-sensor link. Without toe joints, "end-effector link == force-sensor link" is assumed. With toe joints, "end-effector link != force-sensor link" is assumed.

Hands and arms functionality

In AutoBalancer, hands and arms has several functionality for two legs and two arms robot. 1. No ik mode. If startAutoBalancer with ["rleg", "lleg"], arms IK are not solved while MODE_ABC.

2. IK mode without fix. If startAutoBalancer with ["rleg", "lleg", "rarm", "larm"] and leg_names is ["rleg", "lleg"] and is_hand_fix_mode is false, arms IK are solved while MODE_ABC and Hands target pos are determined according to COG velocity during walking. So, hands seems to follow upper body swinging and total behavior looks like No IK mode.

3. IK mode with fix. If startAutoBalancer with ["rleg", "lleg", "rarm", "larm"] and leg_names is ["rleg", "lleg"] and is_hand_fix_mode is true, arms IK are solved while MODE_ABC. Hands target pos Y in foot coords are not change during walking and pos X in foot coords is determined according to COG velocity X. So, hands seems to be fixed and follow foot projected coordinated.

4. Four legged walking mode. If startAutoBalancer with ["rleg", "lleg", "rarm", "larm"] and leg_names is ["rleg", "lleg", "rarm", "larm"], arms IK are solved while MODE_ABC and hand coords are determined by GaitGenerator as walking target.

Data Ports

Input Ports

port name	data type	unit	description
qRef	RTC::TimedDoubleSeq	[rad]	Reference joint angles
zmpIn	RTC::TimedPoint3D	[m]	Input ZMP in base frame
basePosIn	RTC::TimedPoint3D	[m]	Input base position
baseRpyIn	RTC::TimedOrientation3D	[rad]	Input base orientation (RPY)

Output Ports

port name	data type	unit	description
q	RTC::TimedDoubleSeq	[rad]	Output reference joint angles
zmpOut	RTC::TimedPoint3D	[m]	Output ZMP in base frame
basePosOut	RTC::TimedPoint3D	[m]	Output base position
baseRpyOut	RTC::TimedOrientation3D	[rad]	Output base orientation (RPY)
baseTformOut	RTC::TimedDoubleSeq	[m], [rad]	Output base position and orientation (RPY)
accRef	RTC::TimedAcceleration3D	[m/s^2]	Output attitude sensor's acceleration
contactStates	RTC::TimedBooleanSeq	NA	Reference end-effector contact states (True=contact, False=no-contact)
controlSwingSupportTime	RTC::TimedDouble	[s]	Time of swing phase

Service Ports

Service Providers

port name	interface name	service type	IDL	description
AutoBalancerService	service0	AutoBalancerService	OpenHRP::AutoBalancerService	AutoBalancerService

Service Consumers

N/A

Configuration Variables

name	type	unit	default value	description
debugLevel	int		0	debug level

Configuration File

N/A