cmu-robotics

As of 10 June, this file has been rewritten for how to build the different CPs for RR3. Details about test design and current limitations can be found at:

• documents/test-design-cp1.md
• documents/test-design-cp2.md
• documents/test-design-cp3.md

documents contains markdown files with draft descriptions of each challenge problem following the provided format. The APIs describing the interfaces for TA and TH in each CP case are given in documents/swagger-yaml as valid Swagger files---with automatically produced markdown also checked in for convenience.

All of these are currently still under revision and will change through the process as we refine our research goals.

Inside each CP directory, you'll find at least:

• a directory ta/ that's the result of running swagger-codegen on the Swagger definitions and then filling in the resulting stubs to mesh with the underlying subsystems.

• a docker-compose.yml that specifies how to composes our system with the LL TH image

• a docker-compose-no-th.yml that specifies a how to compose a docker container like the one above, but without the TH from Lincoln Labs

Each CP's TA follows the sequence diagram as described in this repo, so it begins by sending a message to /ready on the TH and waiting for a reply. Some of the TAs may have a fallback for reading a ready message JSON object from the filesystem for debugging, but that is not uniform.

DockerHub Containers

For each challenge problem, we have tagged a container for integration week. The containers will allow you to run without being able to build. They are:

• cmumars/cp1:latest
• For CP2 see cp2/ta/README.md
• cmumars/p2-cp3:RR3

To get these, you can simply do docker pull cmumars/p2-cp3:latest for example. You will need to change the docker-compose files to refer to these builds though.

Building CP1

1. Clone the CP1 base repo, git clone [email protected]:cmu-mars/cp1_base.git somewhere on the file system, which we'll call CP1_BASE.

2. Build cmu-mars/base (note that this container is shared with CP3):

   cd CMU_ROBOTICS/mars-main-p2
   docker build -t cmu-mars/base .

3. Build cmu-mars/gazebo (note that this container is shared with CP3):

   cd CMU_ROBOTICS/cp-gazebo-p2
   docker build -t cmu-mars/gazebo .

4. Build cmu-mars/cp1_base:

   cd CP1_BASE
   docker build -t cmu-mars/cp1_base .

5. Build cmu-mars/cp1_rb:

   cd CMU_ROBOTICS/rainbow-p2
   docker build -t cmu-mars/cp1_rb -f Dockerfile-cp1 .

6. Build cmu-mars/cp1:

   cd CMU_ROBOTICS/cp1/ta
   docker build -t cmu-mars/cp1 .

7. Compose cmu-mars/cp1 with the TH:

   cd CMU_ROBOTICS/cp1/ta
   TH_PORT=8081 TA_PORT=8080 docker-compose up

Building CP2

Instructions for building and interacting with CP2 can be found at: cp2/ta/README.md.

Running CP3

1. Pull the docker image cmumars/p2-cp3 from DockerHub
2. In the directory that you are wanting to compose in, ensure that the directorys roslogs, logs exist (these are where logs will be put), and ensure that they are Readable, Writable, and Executable for everyone.
3. Compose with:

TA_PORT=5000 TA_PORT=5001 docker-compose -f docker-compose-mitll-harness.yml up

Notes on limitations for CP3 in RR3

Some of the configurations do not have a plan associated with them on some paths, meaning that an A case will not run. If the TH tries to start one of these tests, it will get an error from the TA. There may also be some of these that exist during the actual evaluation. It is about 5% of the configuration/path space currently. We aim to reduce this.

Building CP3

Note: Due to a last minute disappearance of one of the Unix packages (ros-kinetic-mrpt-localization) it is not possible to build CP3 from scratch. Use the DockerHub version instead. We are hoping this is temporary, but will develop a workaround before evaluation.

1. Clone the CP3 base repo, git clone [email protected]:cmu-mars/cp3_base.git somewhere on the file system, which we'll call CP3_BASE.

2. Build cmu-mars/base (note that this container is shared with CP1):

   cd CMU_ROBOTICS/mars-main-p2
   docker build -t cmu-mars/base .

3. Build cmu-mars/gazebo (note that this container is shared with CP1):

   cd CMU_ROBOTICS/cp-gazebo-p2
   docker build -t cmu-mars/gazebo .

4. Build cmu-mars/cp3_base:

   cd CP3_BASE
   docker build -t cmu-mars/cp3_base .

5. Build cmu-mars/cp3_rb:

   cd CMU_ROBOTICS/rainbow-p2
   docker build -t cmu-mars/cp3_rb -f Dockerfile-cp3 .

6. Build cmu-mars/cp3:

   cd CMU_ROBOTICS/cp3/ta
   docker build -t cmu-mars/cp3 .

7. Compose cmu-mars/cp3 with the TH:

   cd CMU_ROBOTICS/cp3/ta
   TH_PORT=8081 TA_PORT=8080 docker-compose up

Notes on getting this working in Windows 10

To run on Windows 10, you need to make sure that the correct port forwarding is set up. So, as administrator, you may need to run:

netsh interface portproxy add v4tov4 listenaddress=127.0.0.1 listenport=8080 connectaddress=192.168.99.100 connectport=8080

Where listenport and connectport are the TA_PORT specified in docker compose, and connectaddress is the IP of the host machine.