highway-env

Deep Reinforcement Learning based Decision-Making in Autonomous Driving Tasks

Heterogeneous Multi-agent Version of Highway-env

An extension of the Planner-Actor-Reporter framework applied to autonomous vehicles in Highway-Env and CARLA.

Autonomous Driving W/ Deep Reinforcement Learning in Lane Keeping - DDQN and SAC with kinematics/birdview-images

Implementation of Deep Deterministic Policy Gradient (DDPG) method on autonomous vehicle within the highway-env

Reinforcement Learning Final Project

stress testing black-box AVs with MARL

DQN-based autonomous driving agent on Highway-Env. Trained with Stable-Baselines3. Balances speed, collision avoidance, and lane discipline.

🚗 Analyze and visualize decision-making in autonomous driving RL agents using Integrated Gradients for clearer interpretability in complex driving tasks.

Reinforcement Learning : Autonomous parallel parking task. implementing SAC and DreamerV3's World Model on Highway-env

Budgeted counterfactual safety-reversal testing for simplified autonomous-driving simulations — an exploratory, reproducible research toolkit (closeout: INCONCLUSIVE).

Interpretability in Autonomous Driving: Visual Attribution Analysis of RL Agents

Development of autonomous agents driving safely in highway environment using Reinforcement Learning

Risk-Sensitive Distributional Reinforcement Learning (QR-DQN vs. DQN) for Autonomous Driving in High-Density Highway, POMDP, Adversarial Cut-In, and Multi-Agent Platoon Environments

PPO agent trained to navigate highway-env - mean reward 20.7, zero crashes, with reward curve and video evaluation

Benchmark of five open-weight LLMs as driving decision-makers in highway-env, with a rule-based safety layer. Paper under review.

Belief-Aware Guided SAC for autonomous driving under partial observability (arXiv:2605.26155, CVPR 2026 WAD)

using reinforcement learning

Text-to-Reward RL for highway-env: an LLM writes, validates, and evolves complete Python reward functions for a PPO agent, with built-in reward-hacking detection.

A reward-free autonomous driving agent in highway-env that plans by counting survivable futures, exploring the limits of viability theory when survival and task progression diverge.