This project aims to create an algorithm to allow safe, autonomous navigation for a single robot in an unknown environment.

Before running the simulator, make sure install every packages needed for this project. Otherwise you will get a list of errors.

Pedestrian ID are indexed from 1 instead of 0. First, you need to specify some other pedestrian on the command line. For example, run:

python3 Main.py --ped-id-to-replace 3

Question: Why the simulator is looking for a specific pedestrian?

Answer: The specific pedestrian will replace with a robot and then run a navigation algorithm(e.g. DeepMotion and SFM)

To run the simulator, first cd into the directory you cloned this repository to, then run:

./run_once.sh

The above command will generate a lot of frames so that you can visually know how robot naviation works. To start the SVG viewer, you can run:

eog *.svg

There are many settings that can be configured from the command line. To see which flags are available, just run:

python3 Main.py --help

There are some (but not many) unit tests available in the testcode module. These can be run with python3 -m testcode.<test-file-name>. For example, to run the geometry unit tests in testcode/geometry_test.py, you would use:

python3 -m testcode.geometry_test

To train the neural network(s) for the DeepMotion navigation algorithm (DeepPredNavAlgo class), go into the human_action directory and run python3 train_C.py action data/training_human_data.json. The networks will be saved in the dnns folder. The DeepPredNavAlgo class takes a net_load_file parameter that specifies the file to load the network from.

Use tabs, never spaces. Spaces break the Cython compilation. Also, use Unix line endings.