active-target-localization

Codebase for the paper titled "Active Localization of Multiple Targets from Noisy Relative Measurements" (WAFR 2020)

Check out our project page!

Platform and dependencies

• Ubuntu 16.04
• Python 3.6
• Pytorch 1.4.0
• OpenAI gym 0.17.2 (https://github.com/openai/gym)
• Visdom 0.1.8

Training

For training a session, run:

python -m target_localization.train --sess atl --num_targets 2 

Add the flag --image_representation to use the encoding of the the belief-map image as part of the observation function. Refer to the paper for more details.

Testing

For evaluating the agent policy, run:

python -m target_localization.eval --sess atl --num_targets 2