This repository contains code to record and reproduce social navigation trajectories. For now, it works with real recordings only, simulations will be added later.
To reproduce a recorded trajectory:
git clone https://github.com/ljmanso/socialnavigationdatasetdiscussioncd socialnavigationdatasetdiscussionwget https://ljmanso.com/files/sn3_examples.tgztar xzf sn3_examples.tgzsudo docker build -t sn3 .sudo docker run -it --net=host --rm --gpus all -v $(pwd):/project -w /project sn3 bash
That will give you a shell, but we want to connect using a VNC client. In this case we are using xtightvncviewer, but other clients should work just as well. The VNC server uses "sn3" as a password -feel free to change it (run the following outside the docker):
xtightvncviewer localhost:5920- Type the password when asked.
That will show a window manager in the container. To replay a simulated trajectory you can run the following either using an xterm within the virtual desktop or within the docker prompt:
python3 code/sn3_replay.py FILE
where FILE is any of the files in the sn3_examples directory. For instance:
python3 code/sn3_replay.py sn3_examples/1716901588_data.pickle