06/06/2025, 06:45 AM UTC
机器人学会更智能、更快速地打包Robots Learn To Pack Smarter And Faster
➀ 研究人员开发了一种基于任务与运动规划（TAMP）的算法（cuTAMP），利用GPU并行计算同时评估数千种打包方案；
➁ 该方法将决策时间缩短至数秒，可在无需预先训练的情况下实现复杂空间内多样物体的高效装箱；
➂ 经真实机器人及模拟环境验证，该技术具备工业场景适应性与多任务扩展性，未来可结合大语言模型执行高级指令。
➀ Researchers developed a TAMP-based algorithm (cuTAMP) that uses GPU-accelerated parallel computing to evaluate thousands of packing solutions simultaneously;
➁ This method reduces decision-making time to seconds, enabling efficient packing of varied objects in tight spaces without prior training;
➂ Demonstrated in real robots and simulations, the approach shows versatility for industrial tasks and potential integration with AI language models.
---
本文由大语言模型（LLM）生成，旨在为读者提供半导体新闻内容的知识扩展（Beta）。
