A Novel MLLM-Based Approach for Autonomous Driving in Different Weather Conditions
Bibliographic record
Abstract
Autonomous driving (AD) technology promises to revolutionize daily transportation by making it safer, more efficient, and more comfortable. Its role in reducing traffic accidents and improving mobility is vital to the future of intelligent transportation systems. AD systems (ADS) are expected to function reliably across diverse and challenging environments. However, existing solutions often struggle under harsh weather conditions such as foggy, rainy, or stormy circumstances, and mostly rely on unimodal inputs, thus limiting their adaptability and performance. Meanwhile, multimodal large language models (MLLMs) have shown remarkable capabilities in perception, reasoning, and decision-making, yet their application in AD, particularly under extreme environmental conditions, remains largely unexplored. Consequently, this paper proposes MLLM-AD-4o, a novel AD agent that leverages prompt engineering to integrate camera and LiDAR inputs for enhanced perception and control. MLLM-AD-4o dynamically adapts to available sensor modalities and is built upon GPT-4o to ensure contextual reasoning and decision-making. To support realistic evaluation, the agent was developed using the LimSim++ framework, which integrates the SUMO and CARLA driving simulators. Experiments are conducted under harsh conditions, including bad weather, poor visibility, and complex traffic scenarios. The MLLM-AD-4o agent’s robustness and performance are assessed for decision-making, perception, and control. The obtained results demonstrate the agent’s ability to maintain high levels of safety and efficiency, even in extreme conditions, using different perception components (e.g., cameras only, cameras with LiDAR, etc.). Finally, this work provides valuable insights into integrating MLLMs with AD frameworks, paving the way for fully safe ADS.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".