Humanoid Gait Optimization Based on Human Data

Abstract

Original scientific paperAchieving a stable, human-like gait for humanoid robots is a challenging task. While a variety of techniques exist to generate stable walking patterns, only little attention has been paid to the resemblance to the human gait. Popular gaits, for example, apply the strategy to bend the knees and to swing the torso in the lateral direction in order to ensure stability by shifting the center of mass. As a result, the walking patterns do not look very humanlike. However, human resemblance is an important aspect whenever robots are designed to coexist and interact with humans. In this article, we present techniques to optimize a given, stable gait of a humanoid robot with respect to human resemblance. To acquire human data, we use a full-body motion capture system. We propose four different optimization algorithms that work at joint angle basis and use the joint angle difference as measure of similarity. The experiments carried out with a HOAP-2 robot in simulation demonstrate that all techniques generate a gait that is significantly more human-like compared to the robot's initial gate. As the results show, the optimization methods based on hill climbing and policy gradient estimation yield the best performance. Key words: Humanoid robots, Gait optimization, Human-like walkingOptimizacija držanjačovjekolikih robota utemeljena na podacima izčovjekovog hoda. Postizanje stabilnog,čovjekolikog držanjačovjekolikih robota vrlo je zahtjevan zadatak. Iako postoji mnoštvo tehnika koje se koriste za postizanje stabilnih uzoraka hodanja, malo se pažnje pridaje sličnosti s ljudskim držanjem. Primjerice, neke tehnike koriste strategiju savijanja koljena i njihanja torza u lateralnom smjeru kako bi se osigurala stabilnost kroz promjenu položaja centra mase. Kao rezultat toga, uzorci hodanja nisu sličničovjekovom hodu. Ipak, sličnosť covjeku važan je aspekt u slučajevima kada su roboti izvedeni za suživot i interakciju s ljudima. U ovom sučlanku predstavljene tehnike za optimizaciju danih stabilnih držanjačovjekolikog robota s ciljem sličnostičovjeku. Za prikupljanje podataka očovjeku koristi se sustav za snimanje cjelokupnog gibanja tijela. Predložena sučetiri različita algoritma optimizacije koji koriste zajedničku derivaciju kuta kao mjeru sličnosti. Eksperimenti provedeni koristeći HOAP-2 robot pokazuju da svi postupci generiraju držanje koje je značajno više nalikčovjekovom držanju u odnosu na početno držanje robota. Kao što rezultati pokazuju, metode optimizacije koje se zasnivaju na penjanju uzbrdo i estimaciji gradijenta smjera daju najbolje rezultate.

Humanoid Robot Imitation with Pose Similarity Metric Learning

“…Whereas training with separate networks would cause correlation loss among the DOFs. Other imitation researches are mainly dedicated to humanoid gait or walking movements ( [5], [6], [7]). In conclusion, these works have the following limitations:…”

Robust and efficient walking with spring-like legs

“…While it seems that walking on two legs is naturally internalized in humans, it is a challenge to implement this gait into bipedal robots. One way of implementing walking into an artificial system could be performed by copying anthropomorphic data and reproducing human-like leg kinematics (Zielińska 2009, Wehner andBennewitz 2009). The second possibility is to simplify the motion and, thus, simplify the system, which also leads to a walking gait as demonstrated with passive dynamic walkers (McGeer 1990).…”

Adaptive control for wearable robots in human-centered rehabilitation tasks

“…It mainly consists of proportional feedback position controllers with joint angle gait trajectories as input. For trajectory tracking, the reference trajectories can be defined based on mathematical models of normative gait trajectories [21][22][23] or pre-recorded trajectories from healthy individuals [24][25][26][27] or by the movement of the unimpaired limb [28].…”