Since codebase is not available
- Get the LNN Package Running
- Get vanilla RL running - Make it modular so we can itegrate and test easily
- Combine them to integrate external knowledge with RL COntroller
- Initial Test on Pacman
- Lets try safeRL benchmarks? (https://openai.com/blog/safety-gym/)
- Modular Code - Enable easier switching of Gym environments
- Knowledge base generation
- Working code base
- Show results on atleast 1 example (2 in the best case)
- Report with structure hinting towards the final project submission