chasing-targets-gym 0.1.0

Simple gym environment for multiple agents to chase multiple (dumb) targets

Introduction

This is a simple gym environment that sets up a set of robots and targets for them to chase. These targets are dumb, they simply move at a constant speed and bounce off the “limits” of the simulation environment. The intention is that the robots will chase after these targets, and switch to a new target after catching their current one. The targets are “transparent” and robots are free to ignore “avoiding them”, the intention is that they avoid each other. An example of a simulation with robot controller is shown below.

Usage

Since this uses the gymnasium, you can spin an environment up same as any other env, and you can use our optimized planner. A script is included that shows of this planner and environment when you install this library `chasing-targets-example --max-step=500`.

from gymnasium import Env, make
from chasing_targets_gym.planner import Planner

env: Env = make(
    "ChasingTargets-v0",
    render_mode="human",
    n_robots=10,
    n_targets=3,
    robot_radius=0.1,
    max_velocity=0.5,
    target_velocity_std=0.5,
    max_episode_steps=1000,
)

planner = Planner(
    env.get_wrapper_attr("robot_radius"),
    env.get_wrapper_attr("dt"),
    env.get_wrapper_attr("max_velocity"),
)

Installation

Either you can clone and pip install the source, or you can install via pypi. If installing from source seems to stall for no apparent reason, try –no-build-isolation.

git clone https://github.com/5had3z/chasing-targets-gym && cd chasing-targets-gym && pip3 install -e .

Otherwise install pypi package

pip3 install chasing-targets-gym

Some Credit

I was pointed to a basic environment here but it didn’t really match what I wanted, so I made my own based off this.