multi_robot_task_allocation

Optimization-based controllers for multi-robot task exeution, prioritization, and allocation.

task_as_constraints

Basic code associated with the paper Constraint-Driven Coordinated Control of Multi-Robot Systems

lazy_task_allocation

Basic code associated with the paper An Optimal Task Allocation Strategy for Heterogeneous Multi-Robot Systems

resilient_task_allocation

Basic code associated with the paper A Resilient and Energy-Aware Task Allocation Framework for Heterogeneous Multi-Robot Systems

Requirements

Before running any script, run init.m to add the folders of this repository to the MATLAB path.

swarm_sim

This is included as a submobule in the include directory. After cloning this repository, run:

git submodule init
git submodule update

To run the mixed centralized-decentralied code in resilient_task_allocation/mixed_centralized_decentralized, the mqtt_matlab_interface is used to exchange data between two instances of MATLAB running the centralized and decentralied parts of the algorithm at different rates.

Add the main folder of the repository mqtt_matlab_interface to the MATLAB path and initialize the MQTT MATLAB interface as described in its README file
On one instance of MATLAB, run rta_central_unit.m
On a second instance of MATLAB, run rta_robots.m

The script rta_central_unit.m receives the robots' state and runs the centralized, slow, and non-convex MIQP. Every time an optimal solution is obtained, the task allocation is communicated to the script rta_robots.m where the single robots are simulated, and only a convex QP is solved at a faster rate.

CVX and Gurobi

Mixed-integer quadratic programs in the code are solved using CVX with Gurobi solver (http://cvxr.com/cvx/download/), which requires the following initialization steps:

cd lazy_task_allocation/code/lib/
wget http://web.cvxr.com/cvx/cvx-a64.tar.gz
tar -xvzf cvx-a64.tar.gz
rm cvx-a64.tar.gz

Academic licenses

Request adacemic licenses for CVX and Gurobi here:

http://cvxr.com/cvx/academic/ → receive a cvx_license.dat by email
https://www.gurobi.com/downloads/end-user-license-agreement-academic/ → receive a Gurobi license key xxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx

Option 1: With a VPN connection to an institutional network

In the MATLAB command window:

cd lib/cvx/
cvx_grbgetkey xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
cvx_setup /path/to/your/cvx_license.dat

Option 2: Without a VPN connection to an institutional network

Download the Gurobi software from here https://www.gurobi.com/downloads/gurobi-software/
Extract it to your favorite location (/opt/gurobi is advised in order to allow any user to use it not necessarily through CVX and MATLAB)
cd /opt/gurobi/bin/linux64/bin
./grbgetkey xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
Move the license file gurobi.lic to ~/.matlab/cvx_gurobi.lic

Finally, select Gurobi as CVX solver in the MATLAB command window:

cd lib/cvx/
cvx_solver gurobi

wy-hm/multi_robot_task_allocation