logistic_sim

A heaviliy modified version of the patrolling_sim module to solve the multi-item variant of the MAPD (Multi-Agent Pickup and Delivery) problem. This package implements an allocation and coordination scheme based on a token passing algorithm that enables robot to find collision-free paths to solve MAPD instances.

Getting started

This package has been tested on ROS Kinetic and Ubuntu 16.04

Prerequisites

This package uses the ROS navigation stack and the Stage simulator (all available in the ROS repos).

To run the CLI you need tmux.

Compiling

To compile just cd into the repo and run catkin_make, if compilation fails you're probably missing some dependencies.

After compiling the code run source devel/setup.bash to add the workspace to the environment (it is good practice to add this script to your .bashrc file).

Running

There are two ways to launch the simulation, one using tmux (recommended) and another using xterm

tmux

To run a simulation head into the src/logistic_sim directory and run ./tmux_run_exp.sh.

This script uses tmux to organize the terminals of the different nodes so it is recommended to know at least how to switch between windows and panes. The nodes are organized in different windows:

roscore and free terminal
move_base nodes
agents nodes
task planner node

The script lands on the agents window.

If you desire to terminate the simulation prematurely press ctrl-b X (assuming you're using the default tmux prefix key).

xterm

Inside src/logistic_sim directory run ./run_exp.sh to start the simulation. To stop the simulation run ./stop_experiment.sh

Parameters

Inside the script there are many parameters that can be configured:

The ALG parameter sets which agents algorithm to run:

OnlineDCOPAgent implements the online allocation and planning algorithm
OnlineCentralizedAgent is used in conjunction with two different centralized approaches (one with optimal allocation and the other with a greedy strategy)

The TP_NAME parameter sets the task planner:

OnlineDCOPTaskPlanner must be used together with OnlineDCOPAgent
OnlineGlobalTaskPlanner and OnlineGreedyTaskPlanner must be used with OnlineCentralizedAgent and they implements the two approaches aforementioned.

The GEN parameter sets how the tasks are generated:

uniform creates tasks uniformly distributed in terms of delivery locations and demand values
rand creates random tasks
file reads the tasks from the file specified in the MISSIONS_FILE parameter, this file must be contained in the missions directory
null does not create tasks, they must be provided calling the AddMissions service

An example for the AddMissions service: rosservice call /add_missions "[{DSTS:[42]},{DSTS:[26,42]}]" In this example two missions are inserted, one with a single delivery location and one with two delivery locations

The NROBOTS parameter sets the number of robots to run in the simulation, currently only configurations of 2, 4 or 6 robots are defined, but different ones can be made in the params/initial_poses.txt file. This file contains the starting positions of the robots for each map.

In this moment only the icelab_black,icelab_room,grid and model5 maps are working.

RB-KAIROS

If you want to use the kairos simulator the USE_KAIROS_SIM flag must be set to true. To use the kairos simulator its packages must be installed, instructions can be found here.

If you want to use the real robot then the USE_KAIROS flag must be set to true. In this case the ROS_MASTER_URI must be set to the robot's one, because roscore runs on the robot.

Debugging

If the DEBUG flag is set to True the agents and the task planner will run inside gdb. Custom gdb commands can be given to the robots and the taskplanner by creating specific named files inside the src/logistic_sim directory:

For the taskplanner create a commands_taskplanner.txt file
For a single agent create a commands_<agent_id>.txt file, where the agent id can go from 0 to NROBOTS - 1
To send commands to all agents create commands_all.txt file

Astuxx/RMAPD