Humanoid Robot LOLA - Vision Guided Autonomous Multi-Contact Locomotion

In this video we demonstrate fully autonomous multi-contact locomotion for our humanoid robot LOLA. In contrast to our previous multi-contact videos where contact points for the feet and hands had to be specified manually by the user, this time all contacts are autonomously planned by the robot itself based on the perceived environment. The only input by the user is the desired final goal position (a horizontal position and rotation around the vertical axis). The robot then automatically computes a feasible contact sequence (if possible) and connects the discrete poses with kinematically and dynamically feasible trajectories while considering multi-contact effects (external forces applied at the hands of the robot). Through this experiments we demonstrate the coupling of LOLA's new computer vision (Chair for Computer Aided Medical Procedures Augmented Reality, TUM) and walking pattern generation (Chair of Applied Mechanics, TUM) systems. All algorithms run onboard and in real-time. The scene is not known to the robot (it has to detect it on its own). This work is supported by the German Research Foundation (DFG, project number 407378162).

本视频展示了我们的人形机器人（humanoid robot）LOLA的全自主多接触运动。相较于我们此前发布的多接触类视频中需由用户手动指定手脚接触点的方案，本次所有接触点均由机器人基于自身感知到的环境自主规划。用户仅需提供期望的最终目标位置（包含水平位置与绕垂直轴的旋转角度），随后机器人将自动（若可行）计算出可行的接触序列，并将离散位姿通过运动学与动力学可行的轨迹进行连接，同时兼顾多接触效应——即作用于机器人手部的外力。通过本次实验，我们展示了LOLA全新的计算机视觉系统（慕尼黑工业大学TUM计算机辅助医疗程序与增强现实研究室）与步行步态生成系统（慕尼黑工业大学TUM应用力学研究室）的耦合应用。所有算法均在机载端实时运行。机器人对当前场景无先验知识，需自主完成环境检测。本研究得到德国研究基金会（DFG，项目编号407378162）的资助。

搜集汇总

背景与挑战

背景概述

该数据集展示了人形机器人LOLA的视觉引导自主多接触运动，实现了从用户仅输入目标位置到机器人自主规划接触序列和轨迹的完全自主操作。它集成了计算机视觉和行走模式生成系统，所有算法在未知环境中机载实时运行，体现了多系统耦合和实时处理能力。