ADHERENT: Learning Human-like Trajectory Generators for Whole-body Control of Humanoid Robots

Abstract

Human-like trajectory generation and footstep planning represent challenging problems in humanoid robotics. Recently, research in computer graphics investigated machinelearning methods for character animation based on training human-like models directly on motion capture data. Such methods proved effective in virtual environments, mainly focusing on trajectory visualization. This paper presents ADHERENT, a system architecture integrating machine-learning methods used in computer graphics with whole-body control methods employed in robotics to generate and stabilize human-like trajectories for humanoid robots. Leveraging human motion capture locomotion data, ADHERENT yields a general footstep planner, including forward, sideways, and backward walking trajectories that blend smoothly from one to another. Furthermore, at the joint configuration level, ADHERENT computes data-driven wholebody postural reference trajectories coherent with the generated footsteps, thus increasing the human likeness of the resulting robot motion. Extensive validations of the proposed architecture are presented with both simulations and real experiments on the iCub humanoid robot, thus demonstrating ADHERENT to be robust to varying step sizes and walking speeds.

Dynamic Complementarity Conditions and Whole-Body Trajectory Optimization for Humanoid Robot Locomotion

“…IV-I. An interesting future work consists in adopting Reinforcement Learning techniques, like [73], [74] to warm start the optimization problem. In addition, the definition of references can affect the time necessary to find a solution.…”

Online Non-linear Centroidal MPC for Humanoid Robot Locomotion with Step Adjustment

“…To achieve this objective, we may perform a dynamic extension of the system (4) considering the contact forces as a state of the dynamical system and controlling their derivative [46]. Finally, to improve the overall time performance, we are planning to warm start the non-linear optimization problem with the result of a human-like trajectory planner [47].…”

Design, development, and analysis of the lower body of next-generation 3D-printed humanoid research platform: PANDORA

“…The robot's ability to mimic human motion, along with its multimodal communication skills, helps in performing tasks in collaboration with humans. Recent research on humanoid robots involves different fields such as hardware development [1][2][3][4][5], artificial intelligence [6,7], bipedal locomotion [8,9], and teleoperation [10] which all require various of levels of interaction with the surrounding environment. These interdisciplinary research technologies provide an innovative approach, constantly shaping the growth of humanoid robotics.…”

“…The hydraulically actuated humanoid robot, DRC-Atlas [30], by Boston Dynamics was used by teams without their own humanoid as a research platform. After the DRC, other advanced humanoid robots such as E2-DR [31], TALOS [32], Next-Atlas and HRP-5P [2], TESLA Optimus Bot [33], LARMbot 2 [34], Agility Robotics Digit [35], iCub [6,10], and Institute of Human and Machine Cognition (IHMC) Nadia [36] were released. In particular, the LARMbot 2 is an open-source humanoid robot with parallel actuators that primarily uses additive manufacturing for fabricating its structural components, but is only 0.4 m in height which significantly lowers the actuation constraints [34].…”