HOWTO.md

git Setup

Run the application

Run on Laptop

1. Set backend to system in the [DEFAULT] section of config/pepper.config

Run on Robot

1. Set backend to naoqi in the [DEFAULT] section of config/pepper.py
2. Set ip and port in accordance with robot's address in the [pepper.framework.backend.naoqi] of config/pepper.config

Common steps to run the application

1. Build the Docker image for pepper_tensorflow by running

   > docker build -t cltl/pepper_tensorflow . in that repository.

2. Download GraphDB binary (graphdb-free-<version>-dist.zip) into setup/graphdb-docker/lib and update the graphdb version in the setup/docker-compose.yml file to the downloaded version.

3. Run > docker-compose up from the setup/ folder. This starts

   • bamos/openface
   • GraphDB
   • pepper_tensorflow

   Data used by these containers is stored in setup/data and can be reset by removing all subfolders of that folder. Depending on your docker setup make sure the docker containers have enough memory and CPU available.

4. Make sure there is a GraphDB repository named leolani. If not, run > ./setup/setup-graphdb-repo.sh. If this script fails, you may create the repository manually by accessing http://localhost:7200/webapi on your browser. Under Repository Management, use thePOST endpoint to upload the /setup/repo-config.ttl file.

5. Make sure the pepper/ directory is on the python path, e.g. by invoking python with > PYTHON_PATH='path/to/pepper' python path/to/app or by adjusting the python path in your IDE.

6. Start any Application in pepper/apps/..

7. Done

8. Start/Stop Docker by running docker-compose start/stop when stopping/restarting the application.

Pepper Troubleshooting

No Pepper-laptop connection can be established

1. Make sure Pepper and the laptop are on the same network
2. Verify Pepper has access to network (by pressing belly-button)
3. Make sure ip and port are set correctly in the [pepper.framework.backend.naoqi] section of config/pepper.config or config/default.config

Pepper cannot connect to network

1. Connect Pepper to network using ethernet cable
2. Press belly-button to obtain IP and update ip in the [pepper.framework.backend.naoqi] section of config/pepper.config accordingly
3. Go to robot web page (by entering IP in browser)
4. Go to network settings and connect to wifi
   • If unlisted, reboot robot (and wifi). Make sure wifi is online before robot is.
5. Shutdown robot, remove ethernet cable, and boot again. It now should work...

Problems with speech audio

1. Start an application with StatisticsComponent and look at the STT (Speech to Text) activity
2. If no signal (i.e STT [..........]): use_system_microphone is set to false in the [pepper.framework.backend.naoqi] section of config/default.config and not overriden from config/pepper.config 2. Make sure external mic, if used, is switched on and sensitive enough (use OS settings) 3. Make sure Pepper mic, if used, is not broken?
3. If signal is low (i.e STT [|||.......] is below [pepper.framework.sensors.vad.webrtc] -> threshold):
   1. Make sure you talk loud enough (noisy fans in Peppers head make it difficult)
   2. Make sure you talk in the right microphone (i.e. [pepper.framework.backend.naoqi] -> use_system_microphone)
   3. Make sure the external mics volume is high enough!
4. If signal is too high (i.e. STT [||||||||||]) all the time:
   1. Peppers own mics cannot handle very loud/noisy environments, like fairs
   2. Use a microphone attached to the laptop, instead!
      • don't forget to override [pepper.framework.backend.naoqi] -> use_system_microphone in config/pepper.config and set it to True
5. Microphone should process audio at 16 kHz (i.e. Statistics: Mic 16.0 kHz), if not:
   1. Override cam_resolution and/or cam_frame_rate in the [DEFAULT] section of config/pepper.config with lower values in order to meet performance requirements