JSK's use of humanoids

Humanoids
1

Hi

We at JSK, has similar situation as vigar, we had a lot of (un-organized) ROS-Software also used in DRC and has been applied to several robots including open and closed robot.
walking controller → GitHub - fkanehiro/hrpsys-base: Basic RT components and utilities to control robots using OpenRTM
ros integration → GitHub - start-jsk/rtmros_common: OpenRTM - ROS interoperability packages
unfinished tutorials → GitHub - start-jsk/rtmros_tutorials: Tutorials for rtmros packages
robot model → http://www.jvrc.org/en/humanoid/JAXON/agreement.html
foot step planner → jsk_footstep_planner - ROS Wiki

In my opinion, a Humanoids SIG is a very nice idea, as still many software/approaches are not that easy to reuse for different humanoid robots as they probably could.

I agree, and teams like us, it is not easy to switch from existing system to others, but projet like

  • re-create humanoid msgs and rep humanoid_msgs - ROS Wiki, REP 120 -- Coordinate Frames for Humanoid Robots (ROS.org)
  • create visualizer (rviz plugin) for these msgs
  • From our experience of moveit applied to humanoid robot, Redirecting to Google Groups, we need IK interface for trees (improve kdl_kinematics?), add interface for calculate COM, Momentum and others (kdl_kinematics? robot_model? ), balance calculation method? these could be basics for all of software
  • need mathematical library for humanoid calculation, Preview Control, Resolved Momentum Control? Riccati Equation, …
2 Likes
2

Let’s start a new thread to discuss the revival and extentions to humanoid_msgs. It seems like Vigor’s StepTarget.msg has parameters that more closely match what most groups are commonly using.

An RVIz plugin sounds like a great idea! Most packages are using Visualization Markers for footsteps, but that has been pretty limited in my opinion.

As for humanoid kinematics/dynamics, it seems like Simox VirtualRobot is a lightweight and actively maintained library that has these calculations. There is an example ZMP preview controller based on it https://github.com/TheMarex/libbipedal.

3

@k-okada Regarding improving KDL kinematics for trees, are you aware of this extension of KDL from Leuven?

I have been using this library on our robots, and found it pretty useful.

4

That looks like an interesting library. However it’s licensing under the EUPL will mean that a large part of the ROS community will be unwilling to use the package due to strong copyleft. This makes it unsuitable for integration as a core tool.

From: EUPL or GPLv3? A comparison t… | Joinup

  1. Version 1.1 of the licence was published recently. Why this update?
    In response to the comments and reactions received on v.1.0, the Commission provided some minor clarifications to avoid interpretation problems. Fundamentally however, no big difference between v.1.0 and v1.1 exist, only the drafting has been improved without affecting the spirit of the licence.
    A clarification worth mentioning is the modification of article 1 to improve the management of on-line activities such as “Application Service Providers” (ASP) or “Software as a Service” (SaaS). These activities have been more explicitly included in the scope of the copyleft effect: the EUPL could therefore be described as an “affero-like” licence (AGPLv3)