RI: Medium: Combining Optimal and Neuromuscular Controllers for Agile and Robust Humanoid Behavior

RI：中：结合最佳控制器和神经肌肉控制器以实现敏捷和鲁棒的人形行为

• 批准号: 1563807
• 负责人: Hartmut Geyer
• 金额: $ 100万
• 依托单位: Carnegie-Mellon University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2016
• 资助国家: 美国
• 起止时间: 2016-08-01 至 2020-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1563807&HistoricalAwards=false
• 关键词: RI; Medium; Combining; Optimal; Neuromuscular

The goal of making humanoid robots that assist people in the real world a reality requires that these robots can walk reliably in many different environments and terrains. This project leverages strong progress made by the PIs in the DARPA Robotics Challenge to address new techniques for whole-body locomotion of humanoid robots, based on the blending of model-based and parametric control approaches.A central conundrum in humanoid robotics is that manually designed, policy-based controllers are often more robust than model-based optimizing controllers, but model-based controllers more easily generalize to automated behavior generation with direct control of foot or hand placements. This project tests the hypothesis that incorporating policy-based control as either a cost, a constraint, or a replacement within the model-based online optimal control framework produces controllers that combine the advantages of both approaches.

要想让人形机器人在现实世界中帮助人类，就需要这些机器人能够在许多不同的环境和地形中可靠地行走。该项目利用pi在DARPA机器人挑战赛中取得的巨大进展，在混合基于模型和参数控制方法的基础上，解决人形机器人全身运动的新技术。人形机器人的一个核心难题是，手动设计的基于策略的控制器通常比基于模型的优化控制器更鲁棒，但基于模型的控制器更容易推广到通过直接控制脚或手的位置来自动生成行为。本项目检验了这样一个假设：将基于策略的控制作为成本、约束或替换纳入基于模型的在线最优控制框架，产生了结合两种方法优点的控制器。