Dr Markus Buchholz

post-doc @ HWU

ROS 2 Overview

Motion Planning for real applications

https://space.ros.org/

https://www.therobotreport.com/open-robotics-developing-space-ros/

ROS 2. Concept

https://www.nasa.gov/viper/lunar-operations

• Building robots is difficult.�
• Motors (arms, wheels), sensors, SW have to work all together simultaneously and seamlessly to perform a task.

• The solution is Robot Operating System (ROS), free and open source software, (framework) that provides SW architecture - mechanism, communication layer, tools, and interfaces (SW, HW) to develop, simulate, and deploy Real-Time robotic applications.

https://www.ros.org/

ROS 2. Concept

https://arxiv.org/abs/2211.07752

ROS 2 Design

Robot Programming with ROS2

Architecture

Data Flow

Architecture

API Architecture

ROS 2: Pick and Place application

Vision

Path Planner

IK

Motion Control

Logging

Data Flow

Pick and Place app: Architecture

• We encapsulate robot system task into SW node.
• Nodes communicated over DDS (Data Distributed Service) network throughout messages.
• Each node node can message subscribe or publish.

Vision

Path Planner

IK

Motion Control

Logging

Node

DDS

Node

Node

Node

Node

sub

pub

sub

pub

sub

pub

sub

pub

sub

pub

Architecture

subscriber - publisher : for example sensors, data from nodes

https://docs.ros.org/en/humble/index.html

Architecture

�Services : for example one node set status of the robot to TRUE and other node change the flag to FALSE under some conditions - service here is the mechanism!

Architecture

�Actions: For example, the application requests the robot to move to a specific position and receives feedback on progress and the final result.

Action consist of three parts: a goal, feedback, and a result.�

Functionally is similar to service but they provide permanent feedback and can be cancelled.

ROS 2 is framework designed for developing robotic systems. It provides:

• A communication layer (DDS)
• A mechanism (topics, services, action server)
• A standardized build process (using tools like colcon) to manage package dependencies and create modular, reusable components.
• A set of internal packages and tools for core functionalities like robot control, simulation, sensor integration, and more.

​

Motion Planning

2D (mobile robots)

3D (manipulators, drones, underwater robots)

Grid (X,Y) is the representation of agent searching space

Voxel (X,Y,Z) is the representation of agent searching space

https://github.com/markusbuchholz/uav-blue-robotics-path-planners

Motion Planning for AUV

�IK

(analytical)

​

�Low Level controller�(PID,

Feed Forward, MPC)

​

�Path Interpolator

Defines the geometric path from the start to the goal (joint from IK)

Applies spline interpolation to generate smooth intermediate positions for each joint, (without considering time, velocity, or acceleration)

​

​

��Trajectory Generator

Takes the output from the path interpolator and adds time, velocity, acceleration, and jerk constraints.

Ensures smooth motion by creating a time-parameterized trajectory for each joint, ensuring synchronization and real-time execution.

​

​

�Dynamic model

​

�Target

(program - offline programming)

​

Motion Control in Industrial Manipulators*�(ABB, Fanuc, KUKA, Yaskawa, Kawasaki)

*We do NOT use path planners in most industrial applications. We do not have access to detailed kinematics and dynamic models.

Steps to Synchronize Joint Motions (interpolator)

1. Define Start and Goal Configurations.

2. Compute the Required Motion for Each Joint i :

3. Compute Time for Each Joint:

4. Determine the Maximum Motion Time T,�Find the joint that requires the longest time to reach its target

5. Adjust Velocity Profiles for Shorter Motions:

6. To ensure smooth motion, use spline interpolation for each joint over the synchronized time T.

https://medium.com/geekculture/trajectory-generation-in-c-da36521128aa

Motion Planning is HARD

• Number of DOF.
• Constrains (dynamics).
• Time horizon (real-time computation of path often using pseudo Jacobian).
• Uncertainty (dynamic obstacles).

​

​

https://www.youtube.com/watch?v=DFGE3wTGnu8

Non industrial motion planning

https://ompl.kavrakilab.org

Motion planning (mobile robots)

https://docs.nav2.org/

Path smoothing

Point cloud-> voxel

Voxel is mapped using IK and robot constraints to C-Space (configuration space)

https://www.sciencedirect.com/science/article/pii/S2095809916300443

RRT*, A*, PRM�searches for an obstacle-free path in C-space

Rapidly Exploring Random Trees for Dynamic Obstacle Avoidance

RRT in action (C++ implementation)

ROS industrial

https://github.com/markusbuchholz/Rapidly-Exploring-Random-Trees-for-Dynamic-Obstacle-Avoidance-in-ImGui-Cpp

Moveit computes obstacle free path

obstacle_free_path_planner.mp4

MotoROS2 Yaskawa Motoman Interface for ROS

https://github.com/YaskawaEurope/ros2-starter-for-yaskawa-robots

https://github.com/markusbuchholz/ros2_moveit_manipulators

Moveit 2. Motion planner

https://manipulation.csail.mit.edu/

https://drake.mit.edu/

https://underactuated.csail.mit.edu/

ROS 2 helps to solve real challenges

tether

AUV

ASV

HWU Research setup: Motion Control for Tethered Multi-Robot Systems in Marine Environments with Disturbances

30

We introduce motion control framework for

maritime robotics:

�

​

�

​

�Linker (ROV-Boat)

​

�ROS 2 Interface Boat

​

�MPC Boat

​

�odom_zero_service

​

�ROS 2 Interface ROV

​

�PID ROV

​

�DDS

​

�DVL

​

�GPS

​

�Frame Sync

​

� ROV Motion Planner

​

Software architecture

PC_1 (ROV)

�ArduPilot

​

�ArduPilot

​

PC_2(Boat)

32

First, we employ a path planner with A* algorithm and heuristic strategies designed to account for both environmental disturbances and the tether constraints to compute synchronized motion for both vehicles.

Motion around wind turbine and BOP

Performance of our path planner with A* algorithm

34

We evaluated our framework under real conditions.�

Simulation (3x)

35

We evaluated our framework under real conditions.�

Ocean view

Underwater view

Thanks!