Key results
The company
OrionStar
en.orionstar.comAI-powered service robots for delivery, greeting, cleaning, and warehousing.
Result highlights
- 90% interaction accuracy with Gemini integration
The story
A global robotics company serving over 20,000 customers across 60 countries with service units deployed in healthcare, education, and retail settings.
Legacy robots relied on rigid preset commands and required repeated wake words for simple tasks, creating a mechanical user experience. The system's lack of true contextual awareness made it difficult to support the natural, multilingual interactions required for global expansion.
OrionStar integrated Gemini models into its proprietary AgentOS to unify speech, vision, and motion processing into a single multimodal pipeline. The system allows robots to interpret lip movements and visual cues for natural, multi-turn conversations without constant activation phrases. Development teams also leverage the models to generate code for sub-agent management, accelerating software iteration.
Quotes
“运用 Gemini 大模型，我们正在加速实现使命——‘让机器人像人类一样思考、行动和交流。’ 在猎户星空 AgentOS 的支持下，Gemini Live的低延迟技术，为机器人赋予了更流畅的多轮对话和交互能力，融合语音、视觉与动作，让与机器人的对话能够真正地‘流淌’起来。”