Multi-agent Control System Manual
Table Of Contents
- Roboworks
- Roboworks
- Summary
- Roboworks
- Table of contents
- 1. Introduction to Multi-agent Algorithms
- 2. Multi-agent Communications Setup
- 3. Multi-agent Synchronisation Setup
- 4. Multi-agent ROS Package
- Roboworks
- Introduction to multi-agent algorithms
- 1.1 Multi-agent formation algorithms
- 1.2 Obstacle avoidance algorithms
- Roboworks
- multi-agent communication setup
- Multi-agent communication is one of the key steps to complete a multi-robot formation. When the relative positions of multiple robots are unknown, the robots need to share each other's information through communication to facilitate the establishment of connections. ROS distributed architecture and network communications are very powerful. It is not only convenient for inter-process communication, but also for communication between different devices. Through network communication, all nodes can run on any computer. The main tasks such as data processing are completed on the host side. The slave machines are responsible for receiving environmental data collected by various sensors. The host here is the manager that runs the Master node in ROS. The current multi-agent communication framework is through a node manager and a parameter manager to handle communications among multiple robots.
- Roboworks
- 2.1 The steps to set up multi-agent communications
- Roboworks
- Roboworks
- 2.2 Automatic Wifi connection in ROS
- Roboworks
- Roboworks
- Roboworks
- Roboworks
- Roboworks
- Roboworks
- multi-agent synchronisation setup
- 3.1 Successful master/slave network connection
- 3.2 Troubleshooting network dis-connections
- Roboworks
- multi-agent ros package
- 4.1 ROS Package Introduction
- Roboworks
- Roboworks
- Roboworks
- Roboworks
- Roboworks
- Roboworks
- 4.2 Operation Procedure
- Roboworks
www.roboworks.com.au/support
INTRODUCTION TO MULTI-AGENT ALGORITHMS
1.1 Multi-agent formation algorithms
This ROS package presents a typical problem of multi-agents in collaborative control during a formation drive.
This tutorial lays a foundation for future development on this topic. Formation control algorithm refers to an
algorithm that controls multiple agents to form a specific formation to perform a task. Collaboration refers to
the cooperation between multiple agents using a certain constraint relationship to complete a task. Take the
multi-robot formation drive as an example, collaboration means that multiple robots form a desired formation
together. Its essence is a certain mathematical relationship is satisfied between the positions of each robot.
Formation methods are mainly divided into centralized formation control and distributed formation control.
Centralized formation control methods mainly include virtual structure method, graphical theory method and
model predictive method. Distributed formation control methods mainly include leader-follower method,
behaviour-based method and virtual structure method.
This ROS package applies the leader-follower method in distributed formation control method to execute the
multi-robot formation drive. One robot in the formation is designated as the leader, and other robots are
designated as slaves to follow the leader. The algorithm uses the movement trajectory of the leading robot to
set the coordinates to be tracked by the following robots with certain direction and speed. By correcting the
position deviations from the tracking coordinates, the followers eventually will reduce the deviation between the
follower and the expected tracking coordinates to zero in order to achieve the objectives of formation drive. In
this way, the algorithm is relatively less complicated.
1.2 Obstacle avoidance algorithms
A common obstacle avoidance algorithm is the artificial potential field method. The movement of the robot in a
physical environment is regarded as a movement in a virtual artificial force field. The nearest obstacle is
identified by LiDAR. The obstacle provides a repulsive force field to generate repulsion to the robot and the
target point provides a gravitational field to generate gravitational force to the robot. In this way, it controls the
motion of the robot under the combined action of repulsion and attraction.
This ROS package is an improvement based on the artificial potential field method. Firstly, the formation
algorithm calculates the linear and angular velocity of the Slave follower. Then it increases or decreases the
linear and angular velocity according to the obstacle avoidance requirements.When the distance between the
Slave follower and the obstacle is closer, the repulsion force of the obstacle to the Slave follower is greater.
Meanwhile the change of the linear velocity and the angular velocity variations are greater. When the obstacle
is closer to the front of the Slave follower, the repulsion of the obstacle to the Slave follower becomes greater
(the front repulsion is the biggest and the side repulsion is the smallest). As a result, the variations of the linear
velocity and the angular velocity are greater. Through the artificial potential field method, it improves a solution
ROBOWORKS