dev0guy / amgu

🚦 Traffic Management System 🚏 With Deep Reinforcement Learning 🚗

amgu's Issues

Single Entry point

• Add a single script that runs each option
• Add attacks for the future from prev project

Create Package - premade(models)

• CNN
• FCN
• Qeueue
• Random

Feature: Create Simple AutoEncoder

Create :

• State prediction runner
• Future State prediction runner

Cityflow2d env <information_dim,intersection,lanes*divides>

Package- Docs

• docs using pdoc

Simple Api, for both ray and stable-baseline

There is a need to update all of the Api for much easier use and future integration

Second Step- Tensor board

• How and Where to save to data from rllib
• Create simple tensorboard
• Decide on graph to show on tensorboard
• Create tensorboard with those graphs

Create Lib for are product

Currently, the package does not follow the guideline of PyPi package.
Therefore, there is a need to create a project hierarch + docs (easy Api) + abstract all methods for easy access

• project hierarch
• docs
• abstract all methods

Create Package - premade(reward)

• QeueueLength
• DeltaQeueueLength
• AvgWaitingTime

Create Test For All cityflow env

Create Package - premade(runner)

• RayRunner

Create Package- abstract

• model
• preprocessor
• reward
• runner

Create Package - premade(cityflow)

• CFWrapper
• MultiDiscreteCF
• DiscreteCF

First Steps - Init

• db- create database, run and visualise json file.
• create simple custome env for rllib
• create simple custome model for rllib

Create Package - premade(preprocessor)

• Lane Length
• Vanila

Create Test For Traffic

We need to create a simple unitest module for the package.
The function for each file has been created & needs to be implemented.

Create Docs

Docs add been created however there are two env inside the docs on traffic.

Create Simple 1d ENV <action,action_time, lane queue, lane jam>

Package- Add Attacks Moulde

Environment path isn't working

Is your feature request related to a problem? Please describe.
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]

Describe the solution you'd like
A clear and concise description of what you want to happen.

Describe alternatives you've considered
A clear and concise description of any alternative solutions or features you've considered.

Additional context
Add any other context or screenshots about the feature request here.

Add Visuluzation & Eval Of Results

Third Step- Multiagents

• Create New ENV for multiagent
• Multiagent models
• Compare single agent with couple of intersection against multiagent