Robotics framework JITOMA uses LLMs to optimize 3D scene graphs

Researchers have developed JITOMA, a novel framework designed to improve the efficiency of 3D Scene Graphs (3DSGs) in robotics. This system combats perceptual saturation by employing a just-in-time growth process, where only task-relevant information is processed dynamically. JITOMA utilizes a top-down task heatmap to filter observations and activates specific subgraphs based on queries from a backend Large Language Model (LLM). A new benchmark, JITOMA-Bench, has been introduced to evaluate these dynamic capabilities, showing significant reductions in graph size and captioning latency while maintaining stable processing times. AI

IMPACT This framework could enable more efficient and responsive robotic agents by optimizing perception and memory processing through dynamic subgraph activation.

RANK_REASON Academic paper introducing a new framework and benchmark. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• 3D Scene Graphs
• arXiv
• JITOMA
• JITOMA-Bench
• LLM

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →