Arena Educational

Arena-Education Worksheet #1 (For Arena 3.0)

Installation

Follow the automatic installation in the documentation.

First Steps

Try out your first simulation by running

roslaunch arena_bringup start_arena.launch

Answer the following questions:

1. Which simulator is running?
2. Which robot was spawned?
3. How are the obstacle locations determined?

Solution

1. flatland, 2. burger, 3. randomly

Launch Arguments

Learn about the available launch arguments by reviewing the relevant documentation.

Launch a simulation with the following properties:

• the simulator is flatland,
• the robot is jackal is spawned in map_empty,
• the local planner is dwa,
• the robot task mode and obstacle task mode are both scenario.

Solution

roslaunch arena_bringup start_arena.launch model:=jackal local_planner:=dwa map_file:=map_empty tm_robots:=scenario tm_obstacles:=scenario

Manipulating the Simulation

RViz

The most direct way to manipulate the simulation is by using the RViz UI. Utilize all three tools and try to achieve the following:

• Make the robot not reach its goal for a whole 60 seconds.

rosparam

Some launch arguments can be dynamically updated at run time without having to start the simulation. Read the task mode documenation carefully and dynamically set tm_obstacles to random.

Solution

rosparam set tm_obstacles random

Verify that the obstacles are indeed spawned randomly upon each subsequent reset.

task_generator.yaml

All essential simulation parameters are specified in a central configuration file task_generator.yaml. Locate this file in arena_bringup/configs/task_generator.yaml and open it for editing.

Helpful tip: in order to directly get the exact location of arena_bringup, run rospack find arena_bringup

Global Configuration

Within the configuration you will find the parameter timeout which specifies the maximum time in seconds that the robot can take to reach the goal. Try lowering this limit so that the robot doesn’t reach its goal in time.

Task Configuration

Since we are running tm_robots:=scenario and tm_obstacles:=random at this point, the subsections SCENARIO and RANDOM are relevant to our simulation. We can manipulate their parameters at run time.

Change the scenario by setting SCENARIO/file to an applicable scenario you found in arena_bringup/configs/scenarios.

Solution

Any of map_empty_X.json are fine.

Change the number of obstacles spawned by setting RANDOM/<type>/<min|max> to higher or lower values. Go crazy with this one.

Starting a new Simulation

Some simulation properties cannot be changed at runtime and require a full restart of the simulation. Abort the simulation by sending a keyboard interrupt to the console (Ctrl+C) and wait for everything to stop.

Change the scenario to arena_hospital_scmall_1.json in task_generator.yaml.

Run a new simulation with the updated properties:

• the simulator is gazebo,
• the robot burger is spawned on map arena_hospital_small,
• the local planner is teb,
• the robot task mode and obstacle task mode are both scenario.

Solution

roslaunch arena_bringup start_arena.launch model:=burger local_planner:=teb map_file:=arena_hospital_small tm_robots:=scenario tm_obstacles:=scenario

Wait for Gazebo to fully open (the first model indexing might take a while) and verify that both a Gazebo and an RViz window have opened. All the functionalities taught in the previous section still apply to the current simulation. Apply them to the current simulation and furthermore familiarize yourself with the Gazebo controls.

Headless

While using the Gazebo simulator, you might not necessarily want the added overhead of opening the Gazebo UI. Try starting a semi-headless simulation by adding the launch argument headless:=1.

Benchmark

A special mode is provided by the module benchmark. It is used to automatically start a series of simulations and record data along the way for the purpose of benchmarking planners against each other.

Configuration

Navigate to arena_bringup/configs/benchmark and open the file config.yaml for editing. A benchmark consists of a unique combination of

1. contest - a list of planners (contestants) competing against each other
2. suite - a list of challenges (stages) that each contestant is subjected to

Select a contest and a suite from the respective subdirectories and write them to the config.

Launch

Launch the benchmark by running

arena_bringup start_arena.launch tm_modules:=benchmark

The rest is handled fully automatically. Simulators will open and close multiple times, this can lead to many error messages in the console. This is normal and no cause for concern. If you do not need the simulator UIs at all, add the launch argument headless:=2 to your future benchmark launches.

You may notice that the task_generator.yaml is changed - a backup can be found in task_generator.yaml.bkup. Do not touch either file during the benchmark. Once the benchmark finishes, the initial state will automatically be restored.

Follow the progress of the benchmark by reading the relevant log file in the subdirectory logs. Wait for the entire benchmark to finish.

Postprocessing

The raw result of the benchmark, identified by its run_id from the log file, is located in arena_evaluation/data/<run_id> (find using rospack find arena_evaluation). Further postprocessing of the data is provided by the arena_evaluation package and is covered in its own tutorial.