Reducing Latency in LLM-Based Natural Language Commands Processing for Robot Navigation