Gym for 2D active perception

This is a repository for openai gym environment. The general task is navigation of MAV with limited field of view in unknown and dynamic environment.

Installation

To run the code, it would be suggested to use anaconda to setup the environment:

conda create -n myenv --file package-list.txt

Basic Usage

To run an experiment, an example command is listed below:

python main.py --gaze-method Oxford --agent-number 10 --render --map-size 640 640 --map-scale 10 --dt 0.1 --agent-max-speed 20 --agent-radius 15 --drone-max-speed 40 --drone-max-acceleration 15 --drone-radius 10 --drone-max-yaw-speed 180

There are many parameters that can be pre-defined. By running python main.py -h, a list of all parameters can be seen:

optional arguments:
  -h, --help            show this help message and exit
  --gaze-method GAZE_METHOD
                        Method used (default: Oxford)
  --render              render (default: False)
  --dt DT               environment update time (default: 0.1)
  --map-size MAP_SIZE [MAP_SIZE ...]
                        map size (default: 640*480)
  --map-scale MAP_SCALE
                        map compression scale when building gridmap (default:
                        10 i.e. 64*48 gridmap)
  --agent-number AGENT_NUMBER
                        agent number (default: 5)
  --agent-max-speed AGENT_MAX_SPEED
                        agent max speed (default: 20)
  --agent-radius AGENT_RADIUS
                        agent radius (default: 10)
  --drone-max-speed DRONE_MAX_SPEED
                        drone max speed (default: 40)
  --drone-max-acceleration DRONE_MAX_ACCELERATION
                        drone max acc (default: 15)
  --drone-radius DRONE_RADIUS
                        drone radius (default: 10)
  --drone-max-yaw-speed DRONE_MAX_YAW_SPEED
                        drone max yaw speed (default: 80 degrees / s)