Episodic non-Markov Localization

Implementation in C++ of Episodic non-Markov Localization [pdf]. Please cite this work using the following bibtex:

@article{biswas2016episodic,
  author = { Joydeep Biswas and Manuela M. Veloso },
  url = { https://www.joydeepb.com/Publications/ras_episodic_nonmarkov_localization.pdf },
  doi = { 10.1016/j.robot.2016.09.005 },
  issn = { 0921-8890 },
  year = { 2017 },
  pages = { 162 - 176 },
  volume = { 87 },
  journal = { Robotics and Autonomous Systems  },
  title = { Episodic non-Markov localization },
}

ROS Dependencies

1. ROS
2. AMRL Maps
3. AMRL ROS Messages

Compiling

1. Run ./InstallPackages to install the dependencies on *ubuntu >=14.04 .

2. Add the working directory to the ROS_PACKAGE_PATH environment variable with:

    export ROS_PACKAGE_PATH=`pwd`:$ROS_PACKAGE_PATH
   

3. Install amrl_msgs and add it to the ROS_PACKAGE_PATH variable.

4. run git submodule update --init --recursive to pull submodules

5. Run make

Configuration

Configuration files are written in Lua. If enml is launched without any options, it will try to load the robot config file config/robot.lua. There are several example robot configurations, including cobot.lua, ut_jackal.lua, and ut_automata.lua. To specify a different robot config file, use the -r flag:

./bin/enml -r ut_jackal.lua

The base configuration directory is assumed to be config, but it can be overriden using the -c flag. For example:

./bin/enml -c ~/robot_config -r robot1.lua

This will load the ~/robot_config/robot1.lua file.

The robot configuration file defined the ROS topics to listen to, initialization conditions, and EnML algorithm parameters.

Usage

To run enml, simply run ./bin/enml

Visualization

While running, you may want to visualize what's going on. To do this, add the -d1 flag to the run command, which will cause enml to publish visualization messages while running.

To see the resultant visualization, there are 2 options:

VectorDisplay

Obtain and setup the ut-amrl/vector_display repo. Then run ./bin/vector_display, optionally with the map-name parameter matching the map used for enml localization. This will automatically listen to the same topics enml is publishing, and you will see the visualization in the localization gui's window.

WebRViz

Obtain and setup the ut-amrl/robofleet-deployment-server repo.

Run the websocket ./bin/websocket.

Open the webrviz html file in browser, and connect to localhost. This will automatically listen to the same topics enml is publishing, and you will see the visualization in the web rviz's viewport.

Example

1. Download the example ROS bag file of the UT-Jackal navigating from GDC to AHG: 2020-06-03-18-51-39.bag
2. Edit config/robot.lua to match the UT Jackal's configuration:

   RobotConfig = {
     name = "ut-jackal";
     scan_topic = "/scan";
     pointcloud_topic = "";
     odometry_topic = "/jackal_velocity_controller/odom";
     initialpose_topic = "/initialpose";
   };
   

3. Edit the config/enml.lua to set the initial pose in the ut-jackal section:

   ...
   if RobotConfig.name=="ut-jackal" then
     ...
     -- 2020-06-03-18-51-39.bag
     enml.map_name = "UT_Campus";
     enml.starting_loc_x = 131;
     enml.starting_loc_y = -245;
     enml.starting_angle = deg2rad(-85);
   ...
   

4. Run vector_display
5. Run Enml in offline bag replay mode with the example bag file:

   ./bin/enml -d1 -b 2020-06-03-18-51-39.bag