Improved Safety Control for Autonomous Driving using GPT (LLMs) - Case Study

Problem Statement

An open-source platform, Carla Simulator that facilitates tackling the complexities of autonomous driving by enhancing safety and scene perception. Autonomous driving safety is hindered by unexpected behaviors in dense traffic. Enhanced Safety control using multi-domain skilled GPT to make decision for better decision making process. drawing

Designed LLM cum Autonomous Driving Pipeline

drawing

Simulator Representation

Results

Unexpected Overtaking - Rash Drivers

Smart Lane Change Actions for taking Free-Right

Environment Aware Decision making for Collision Avoidance

Overtaking Obstacle

Challenges & Future Research Work

Uncertain Decision Making at Long Curved Track

Lacking Long Term Decision considering time taken to cross the intersection

Reference

  • For Capstone project Poster, Click here
  • Baseline project’s repository, Interfuser
  • For Linkedin Post, Click here
Gptcarla | @narendhiran2000