arXiv:2304.10891v3 Announce Type: replace Abstract: Transformer-based models are becoming a central paradigm in autonomous driving because they can capture long-range spatial dependencies, multi-agent interactions, and multimodal context across perception, prediction, and planning. At the same time, their deployment in real vehicles remains difficult because high-capacity attention-based architectures impose substantial latency, memory, and energy overhead. This survey reviews representative Transformer-based autonomous driving models and organizes them by task role, sensing configuration, and
The survey addresses the critical challenge of deploying advanced transformer models in autonomous driving, a field rapidly progressing from research to commercial application.
This highlights the immediate technical hurdles for scaling autonomous vehicle deployment, particularly regarding computational efficiency and hardware constraints, impacting the timeline and feasibility of widespread adoption.
The focus shifts from purely algorithmic advancement to practical deployment considerations, emphasizing the need for hardware-efficient AI architectures for real-world autonomous systems.
- · AI hardware manufacturers
- · Automotive OEMs investing in efficient AI
- · Specialized AI compression firms
- · Companies relying solely on unoptimized large models
- · Legacy automotive suppliers
Increased investment and R&D into AI model compression and efficient inference for edge devices will follow.
This could accelerate the commercial deployment of higher-level autonomous driving features, provided these efficiency challenges are met.
Successful optimization paradigms developed for autonomous driving could transfer to other latency-sensitive edge AI applications, driving broader AI adoption.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG