Adaptive Walking through Multi-Contact Stabilization and Usage of Partial Contacts for Humanoid Robots

在过去的几十年里，机器人的社会影响力稳步增长。虽然制造业使用非常专业和高效的系统，但不同的应用领域都是类人机器人的目标，即通过模仿人类人类学设计的机器人。潜在的未来应用是灾区救援行动，家庭医疗保健场景和娱乐行业。虽然类人机器人经常出现在文学和媒体中，但今天即使是最复杂的系统也无法跟上人类的技能。这也适用于实验室以外的非结构化环境中的运动，例如在灾区或人群中。灵活性，即使其行为适应先前未知和不断变化的环境的能力，被证明是一项艰巨的任务。该项目旨在改善类人机器人在这种环境中的运动。为此，主要关注的是用于稳定的主动臂使用，例如抵靠墙壁支撑，以及部分脚/臂接触，例如在爬具有狭窄台阶的楼梯时。整个问题分为感知和环境建模、导航和运动规划以及稳定控制等领域。这些领域被划分为子问题，由慕尼黑技术大学的应用力学主席和计算机辅助医疗程序与增强现实主席合作研究。紧密的合作允许联合收割机结合两个小组的专业知识，以实现一个共同的目标，即提高人形机器人的运动能力。考虑到实际应用是一个长期的目标，特别关注的是高效，实时的算法。

Within the last decades the social impact of robots increased steadily. While the manufacturing industry uses extremely specialized and highly efficient systems, different fields of application are targeted by humanoids, i.e. robots designed by mimicking the human anthropology. Potential future applications are rescue operations in disaster areas, home health care scenarios and the entertainment industry. Although humanoid robots often appear quite capable in literature and media, today even the most sophisticated systems can not keep up with the skills of a human being. This also holds for locomotion in unstructured environments outside of the laboratory, e.g. in disaster areas or in the middle of a crowd of humans. The flexibility, i.e. the ability to adapt its behavior to previously unknown and changing environments, proves to be a difficult task. This project aims to improve the locomotion of humanoid robots within such environments. For this purpose especially active arm usage for stabilization, e.g. supporting against walls, and partial foot/arm contact, e.g. while climbing stairs with narrow steps, is in the main focus. The overall problem splits into the areas perception and environment modeling, navigation and motion planning, as well as stabilization control. These areas are divided into subproblems, which are investigated cooperatively by the Chair of Applied Mechanics and the Chair for Computer Aided Medical Procedures & Augmented Reality of the Technical University of Munich. The tight cooperation allows to combine the expertise of both groups in order to achieve a common goal, namely improve locomotion capabilities of humanoid robots. Considering the practical application as a long term goal, special focus lies on efficient, real-time algorithms.