| Nahyung Kim | SeongKyu Choi | Sun Choi | Yejun Lee | YoungJae Cheong |
|---|---|---|---|---|
 |
 |
 |
 |
 |
- Paper
- GPT-4off : On-Board Traversable Probability Estimation for Off-Road via GPT Knowledge Distillation
- Abstract
- This paper proposes a framework for predicting traversable probability in off-road environments by distilling knowledge from large language models (LLMs) such as GPT-4o into lightweight models. The GPT-4off approach utilizes GPT-generated data to train a compact model capable of real-time operation on edge devices such as the NVIDIA Orin board. Unlike traditional systems that focus on identifying traversable areas, this study emphasizes predicting traversable probability, facilitating faster decision-making in complex environments. This is particularly advantageous for unmanned ground vehicles (UGVs), where obstacles and terrain variability present significant challenges. The GPT-4off framework enhances real-time performance through knowledge distillation and domain-specific optimization, ensuring efficient resource use while maintaining LLM-level performance. Experimental results on the RUGD off-road dataset show that the lightweight model achieves GPT-level performance while being deployable on edge devices. This framework effectively reduces human annotation costs and RAM power consumption, improving the practicality of off-road autonomous driving systems and demonstrating the potential to leverage LLM capabilities for low-power, real-time applications.
- RUGD Dataset
- Prompt
- Assume that a military armored vehicle is driving on this path. The armored vehicle can push through all obstacles even without a road, can climb inclines and rough rocks, can pass through small puddles, can move forward without getting stuck in sand, but cannot hit people. \newline Does this photo seem drivable? On a scale of 0% to 100%, what would be the percentage of drivable probability? Please write the final drivable probability(%) in the first line of your answer. Provide three reasons for your estimation for selecting that certain percentage.
- RUGD probaility output
- label.txt
RUGD_final/
└── label.txt
# label.txt
trail_00051.png 80 # [image_name] [probability percentage(%)]
trail_01761.png 70
...
$ npm start