Deep Sensorimotor Control by Imitating Predictive Models of Human Motion

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Singh, Himanshu Gaurav, Abbeel, Pieter, Malik, Jitendra, Loquercio, Antonio
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909752906219520
author Singh, Himanshu Gaurav
Abbeel, Pieter
Malik, Jitendra
Loquercio, Antonio
author_facet Singh, Himanshu Gaurav
Abbeel, Pieter
Malik, Jitendra
Loquercio, Antonio
contents As the embodiment gap between a robot and a human narrows, new opportunities arise to leverage datasets of humans interacting with their surroundings for robot learning. We propose a novel technique for training sensorimotor policies with reinforcement learning by imitating predictive models of human motions. Our key insight is that the motion of keypoints on human-inspired robot end-effectors closely mirrors the motion of corresponding human body keypoints. This enables us to use a model trained to predict future motion on human data \emph{zero-shot} on robot data. We train sensorimotor policies to track the predictions of such a model, conditioned on a history of past robot states, while optimizing a relatively sparse task reward. This approach entirely bypasses gradient-based kinematic retargeting and adversarial losses, which limit existing methods from fully leveraging the scale and diversity of modern human-scene interaction datasets. Empirically, we find that our approach can work across robots and tasks, outperforming existing baselines by a large margin. In addition, we find that tracking a human motion model can substitute for carefully designed dense rewards and curricula in manipulation tasks. Code, data and qualitative results available at https://jirl-upenn.github.io/track_reward/.
format Preprint
id arxiv_https___arxiv_org_abs_2508_18691
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Deep Sensorimotor Control by Imitating Predictive Models of Human Motion
Singh, Himanshu Gaurav
Abbeel, Pieter
Malik, Jitendra
Loquercio, Antonio
Robotics
As the embodiment gap between a robot and a human narrows, new opportunities arise to leverage datasets of humans interacting with their surroundings for robot learning. We propose a novel technique for training sensorimotor policies with reinforcement learning by imitating predictive models of human motions. Our key insight is that the motion of keypoints on human-inspired robot end-effectors closely mirrors the motion of corresponding human body keypoints. This enables us to use a model trained to predict future motion on human data \emph{zero-shot} on robot data. We train sensorimotor policies to track the predictions of such a model, conditioned on a history of past robot states, while optimizing a relatively sparse task reward. This approach entirely bypasses gradient-based kinematic retargeting and adversarial losses, which limit existing methods from fully leveraging the scale and diversity of modern human-scene interaction datasets. Empirically, we find that our approach can work across robots and tasks, outperforming existing baselines by a large margin. In addition, we find that tracking a human motion model can substitute for carefully designed dense rewards and curricula in manipulation tasks. Code, data and qualitative results available at https://jirl-upenn.github.io/track_reward/.
title Deep Sensorimotor Control by Imitating Predictive Models of Human Motion
topic Robotics
url https://arxiv.org/abs/2508.18691