Xplainable RL

Self Driving Vehicles

XRL: Explainable Reinforcement Learning for Autonomous Driving

Problem: Going from Planning to Control [(Text based) Challenge] in a vision language model.

  • VLAM = VLM + A (action using DL/RL framework)
  • Running optimized VLMs on edge deivces.

Requirement: How to obtain an interpretable navigation policy for intelligent driving agents trained using world model with explainable RL for autonomous driving in real world setting (at action level).

  • Training agents on World Models (Internal WM) using RL/IL.
  • Existing explainable AI (XAI): Vision Language Action Models (VLAMs) techniques
  • does not explain the underlying “action” driving policy for agents behavior.
  • Require explainable RL for interpretable features, policy and learning process

Why?

  • Current: Perception architectures based on explainable machine learning.
  • Transition: LLMs (Large Language Models) → LWMs (Large World Models)
  • Future: Cognitive Architectures based on explainable reinforcement learning.

To give your project a background in the portfolio page, just add the img tag to the front matter like so:

---
layout: page
title: project
description: a project with a background image
img: /assets/img/12.jpg
---
Caption photos easily. On the left, a road goes through a tunnel. Middle, leaves artistically fall in a hipster photoshoot. Right, in another hipster photoshoot, a lumberjack grasps a handful of pine needles.
Why XRL?

Explainable reinforcement learning is an emerging subfield of explainable ML. The goal of XRL is to elucidate the decision-making process of reinforcement learning (RL) agents in sequential decision-making settings. Explainable RL Review Milani, Fie Fang CMU,2024) To understand: What the agents will do and why. A novel taxonomy for organizing the XRL. Three high-level categories:

  • ● Feature Importance,
  • ● Learning process and Markov decision process, (LPM) and
  • ● Policy-level (PL)

Why? RL is combined with the generalization and representational power of deep neural networks, which is often required to achieve the desired performance on these tasks.

You can also have artistically styled 2/3 + 1/3 images, like these.

The code is simple. Just wrap your images with <div class="col-sm"> and place them inside <div class="row"> (read more about the Bootstrap Grid system). To make images responsive, add img-fluid class to each; for rounded corners and shadows use rounded and z-depth-1 classes. Here’s the code for the last row of images above:

<div class="row justify-content-sm-center">
  <div class="col-sm-8 mt-3 mt-md-0">
    {% include figure.liquid path="assets/img/6.jpg" title="example image" class="img-fluid rounded z-depth-1" %}
  </div>
  <div class="col-sm-4 mt-3 mt-md-0">
    {% include figure.liquid path="assets/img/11.jpg" title="example image" class="img-fluid rounded z-depth-1" %}
  </div>
</div>