PyTorch implementation of some reinforcement learning algorithms: A2C, PPO, Behavioral Cloning from Observation (BCO), GAIL.

An end-to-end (E2E) reinforcement learning model for autonomous vehicle collision avoidance in the CARLA simulator, using a recurrent PPO algorithm for dynamic control. The model processes RGB camera inputs to make real-time acceleration and steering decisions.

AI agents for the boardgame Splendor

Repository for "Dynamic Multi-Pair Trading Strategy in Cryptocurrency Markets with Deep Reinforcement Learning" (Lebiedź and Ślepaczuk, 2026).

Solving the Atari Breakout environment using Stable Baselines

Unofficial Stable-Baselines3 extension package with MaskableRecurrentPPO.

Completed reproducible Pokémon Red study: 8.24M self-generated actions reached Route 1, but frozen evaluation found zero durable skills.

Completed recurrent-PPO study of cumulative visual learning, held-out evaluation, retention, and rare failure tails in procedural MiniGrid quests.

Implementation of recurrent Meta-Reinforcement Learning (Meta-RL) agents for investigating planning and memory representations learned by the agents.

Reinforcement learning toolkit for training agents to play Pokémon Red. Built on PyBoy, Stable-Baselines3, and Gymnasium. Includes pixel/symbolic/hybrid observation treatments and rliable-based statistical analysis.