ABot-AgentOS: A General Robotic Agent OS with Lifelong Multimodal Memory

ABot-AgentOS adds a deliberative runtime above low-level robotic controllers, combining scene-conditioned planning, isolated skill execution, multi-stage verification, multimodal graph memory, and edge-cloud collaboration. The paper also introduces EmbodiedWorldBench, an executable benchmark covering 16 indoor, outdoor, and hybrid scenes, four difficulty levels, and more than 200 tasks. On reported memory benchmarks, the static system scores 87.5 on LoCoMo, 59.9 on OpenEQA EM-EQA, 88.6 on Mem-Gallery, and 76.5 Acc@All on NExT-QA. A failure-driven self-evolution loop raises three of these scores further.