DriveCode: Domain Specific Numerical Encoding for LLM-Based Autonomous Driving
This work addresses a fundamental barrier for deploying LLM-based autonomous driving systems by improving numerical precision and efficiency, though it is incremental as it builds on existing multimodal LLM approaches.
The paper tackles the problem of numerical reasoning in LLM-based autonomous driving by introducing DriveCode, a novel numerical encoding method that replaces discrete tokens with dedicated embeddings, resulting in superior performance on trajectory prediction and control signal generation across multiple datasets.
Large language models (LLMs) have shown great promise for autonomous driving. However, discretizing numbers into tokens limits precise numerical reasoning, fails to reflect the positional significance of digits in the training objective, and makes it difficult to achieve both decoding efficiency and numerical precision. These limitations affect both the processing of sensor measurements and the generation of precise control commands, creating a fundamental barrier for deploying LLM-based autonomous driving systems. In this paper, we introduce DriveCode, a novel numerical encoding method that represents numbers as dedicated embeddings rather than discrete text tokens. DriveCode employs a number projector to map numbers into the language model's hidden space, enabling seamless integration with visual and textual features in a unified multimodal sequence. Evaluated on OmniDrive, DriveGPT4, and DriveGPT4-V2 datasets, DriveCode demonstrates superior performance in trajectory prediction and control signal generation, confirming its effectiveness for LLM-based autonomous driving systems.