Driving · 2026-10-02

Experience-Based Feasibility-Aware Generative Adversarial Imitation from Observation under Embodiment Mismatch

Yoshiki Takebayashi, Giovanni Perantoni, Hikaru Sasaki, Matteo Saveriano, Takamitsu Matsubara

arXiv:2610.01171PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

EF-GAIfO is experience-based feasibility-aware imitation from observation for state-only demonstrations when human motions may be infeasible for the robot. It trains a state-transition reconstruction model on the robot’s own experience to estimate feasibility via reconstruction error, then uses feasibility weights in the GAIfO discriminator objective. Feasibility estimates become more discriminative as robot experience accumulates. Validated in simulation locomotion and on a real Unitree Go2 quadruped object-reaching-and-picking task.

Why it matters

Robot-free interfaces often provide state trajectories without action labels, motivating imitation from observation. But embodiment/dynamics mismatch can make demonstrated state transitions infeasible, potentially degrading task performance. EF-GAIfO estimates demonstration feasibility from what the robot has experienced, avoiding explicit dynamics models or prior exploration datasets, and progressively expands the feasible region during policy learning.

Method

  • Train a state-transition reconstruction model R on robot experience collected during GAIfO-based policy learning; evaluate demonstration transitions by reconstruction error e(z, R(z)).
  • Convert reconstruction error to a feasibility weight f_w(z) in [0,1] using thresholds τ_low and τ_high.
  • In the GAIfO discriminator objective, weight demonstration samples by f_w to preferentially imitate demonstration transitions supported by robot experience.

Limitation

The authors state that their feasibility estimator mainly captures embodiment-related motion feasibility, and in real-world tasks feasibility can also depend on surrounding environment (terrain, obstacles, object placement, contact conditions), requiring extending the estimator to condition feasibility on environmental context.

Abstract (from arXiv)

With the increasing use of robot-free demonstration interfaces that provide state trajectories without action labels, imitation from observation has become a promising approach for learning robot behaviors from human demonstrations. However, due to differences in embodiment and dynamics between humans and robots, demonstrated human motions may not be feasible for the robot, potentially degrading policy performance. In this study, we propose Experience-Based Feasibility-Aware Generative Adversarial Imitation from Observation (EF-GAIfO), which estimates the feasibility of state-only demonstrations from the robot's own experience rather than relying on explicit dynamics models or large prior exploration datasets. A key feature of EF-GAIfO is that the notion of feasibility evolves with policy learning: as the policy improves and the robot experiences a broader range of state transitions, the feasible region is progressively expanded, allowing additional demonstrations to be incorporated into learning. This enables feasibility-aware imitation that adapts to the current stage of policy learning, rather than relying on a pre-designed feasibility criterion. We validate the effectiveness of EF-GAIfO on a locomotion task in simulation and on a real quadruped robot performing a object-reaching-and-grasping task.

Related papers

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-10-01 · Indexed by Pub-AI: 2026-10-02 UTC

preprint

Experience-Based Feasibility-Aware Generative Adversarial Imitation from Observation under Embodiment Mismatch

Yoshiki Takebayashi, Giovanni Perantoni, Hikaru Sasaki, Matteo Saveriano, Takamitsu Matsubara

Preprint posted: 2026-10-01 · Original record ↗

arxiv
2610.01171
Explicit source versions (1)
  • v1 · 2026-10-01
  1. Preprint available

    Source date: 2026-10-01 · First detected: 2026-10-25 UTC · Historical baseline

    Source evidence (1)
Abstract · arxiv

With the increasing use of robot-free demonstration interfaces that provide state trajectories without action labels, imitation from observation has become a promising approach for learning robot behaviors from human demonstrations. However, due to differences in embodiment and dynamics between humans and robots, demonstrated human motions may not be feasible for the robot, potentially degrading policy performance. In this study, we propose Experience-Based Feasibility-Aware Generative Adversarial Imitation from Observation (EF-GAIfO), which estimates the feasibility of state-only demonstrations from the robot's own experience rather than relying on explicit dynamics models or large prior exploration datasets. A key feature of EF-GAIfO is that the notion of feasibility evolves with policy learning: as the policy improves and the robot experiences a broader range of state transitions, the feasible region is progressively expanded, allowing additional demonstrations to be incorporated into learning. This enables feasibility-aware imitation that adapts to the current stage of policy learning, rather than relying on a pre-designed feasibility criterion. We validate the effectiveness of EF-GAIfO on a locomotion task in simulation and on a real quadruped robot performing a object-reaching-and-grasping task.