cs.ROSep 7, 2026

Predictive-Coding-Based Autonomous Regulation of Internally Generated and Externally Coupled Processing in Human-Robot Interaction

Authors: Henrique Oyama, Hiroki Sawada, Jun Tani

Organizations: Cognitive Neurorobotics Research Unit Okinawa Institute of Science and Technology Graduate University Okinawa

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Predictive coding characterizes adaptive behavior as a dynamic balance between internally generated predictions and external sensory evidence, yet how an embodied cognitive system can regulate this balance online during ongoing interaction remains poorly understood. This study proposes a predictive-coding-based mechanism for regulating internally generated and externally coupled processing during physical human--robot interaction. The framework employs a predictive-coding-inspired variational recurrent neural network (PV-RNN), in which a meta-prior controls the degree to which posterior inference is constrained by learned prior dynamics. We extend this architecture with an online mechanism that uses reconstruction error accumulated over recent interaction history to select between predefined meta-prior regimes. The mechanism was evaluated across three physical human--robot interaction tasks involving fixed structured, changing structured, and less-constrained interaction. Across all tasks, lower meta-prior values produced the expected increase in posterior--prior divergence and reduction in reconstruction error. More importantly, reconstruction-history-driven regime selection was also associated with reduced prospective prediction error and robot-side physical interaction conflict, demonstrating consequences beyond the retrospective reconstruction objective itself. Task~3 further showed that recent sensory observations can be successfully accommodated while subsequent human motion still departs from the model's prior-generated future trajectory. Overall, these findings show that accumulated reconstruction mismatch can provide an endogenous signal for regulating how strongly subsequent inference relies on learned internal dynamics relative to ongoing sensory input during embodied interaction.

Figures & tables

Figure 1: Hierarchical two-layer PV-RNN architecture with online meta-prior modulation based on temporally accumulated reconstruction error. Solid blue lines represent the generative process, while dotted red lines indicate posterior inference. The shaded region schematically illustrates the sliding inference window used to compute reconstruction error Eˉr,t\bar{E}_{r,t} ; the experiments use TW=100T_{W}=100 . The accumulated reconstruction error regulates probabilistic transitions between internally generated and externally coupled processing.
Figure 2: Control architecture for physical human–robot interaction (adapted from [ 19 ] ). The PV-RNN generates target joint trajectories while force-feedback control accommodates externally imposed human motion, enabling continuous closed-loop interaction between generative dynamics and sensory input. See [ 19 ] for a detailed description of the control architecture and notation.
Figure 3: Probabilistic finite state machine used to generate the structured training sequences for Tasks 1 and 2 (adapted from [ 19 ] ). The four states correspond to movement primitives A–D, with transitions governed by the indicated probabilities.
Layer#d\#d#z\#zτl\tau^{l}wtrw^{\mathrm{tr}}
Layer 2 (Top)30460.01
Layer 1 (Bottom)60230.01
Table 1: PV-RNN training parameters.
Figure 4: Online meta-prior regulation and across-model comparison during Task 1. Left: temporal dynamics for one representative model realization, showing the meta-prior, temporally averaged reconstruction error, and excess torque; shaded regions indicate high- ww (green) and low- ww (orange) processing. Right: across-model comparisons of reconstruction error, excess torque, raw KL divergence in Layers 1 and 2, and prospective prediction error under high and low ww . Bars indicate across-model means, error bars indicate 95% confidence intervals, dots represent independently trained models, and connecting lines indicate paired observations from the same model.
Figure 5: Online meta-prior regulation and across-model comparison during Task 2. Left: temporal dynamics for one representative model realization, showing the meta-prior, temporally averaged reconstruction error, and excess torque; shaded regions indicate high- ww (green) and low- ww (orange) processing. Right: across-model comparisons of reconstruction error, excess torque, raw KL divergence in Layers 1 and 2, and prospective prediction error under high and low ww . Bars indicate across-model means, error bars indicate 95% confidence intervals, dots represent independently trained models, and connecting lines indicate paired observations from the same model.
Figure 6: Online meta-prior regulation and across-model comparison during Task 3. Left: temporal dynamics for one representative model realization, showing the meta-prior, temporally averaged reconstruction error, and excess torque; shaded regions indicate high- ww (green) and low- ww (orange) processing. Right: across-model comparisons of reconstruction error, excess torque, raw KL divergence in Layers 1 and 2, and prospective prediction error under high and low ww . Bars indicate across-model means, error bars indicate 95% confidence intervals, dots represent independently trained models, and connecting lines indicate paired observations from the same model.
Figure 7: Relationship between prospective prediction and physical interaction during Task 3. Less-constrained human motion can depart from the robot’s prior-generated future trajectory even when recent sensory observations are successfully accommodated through posterior inference.
Figure 8: Kernel PCA representations of latent and output dynamics across the three interaction tasks for representative PV-RNN models. Top: deterministic latent dynamics plotted as KPC2 versus KPC1. Bottom: generated output dynamics plotted as the one-step change ΔKPC1\Delta\mathrm{KPC1} versus KPC1. Tasks 1 and 2 share a common kernel-PCA embedding, permitting direct comparison of their structured dynamical organization, whereas Task 3 was embedded separately and exhibits a broader, less discretely organized interaction-driven structure. Green and orange samples indicate high- and low- ww processing, respectively. The remaining independently trained models exhibited qualitatively similar organization.
More in this topic

Predictive Coding

Predictive coding explains perception and learning through predictions, prediction errors, and updates to internal models.

See all 59 papers

Explore similar work

CardsList
Jun 6, 2026

Toward Interaction Dynamics: A Predictive Framework for Safe Physical Human Robot Interaction

Physical human-robot interaction requires yielding transiently to contact yet recovering the commanded reference under sustained load. Finite-stiffness impedance control retains a static deflection there, while predictive alternatives typically optimize a nonlinear robot or impedance model online. Operational-space cancellation instead exposes a translational error double integrator with a fixed transition matrix and a configuration-scheduled input map, making interaction a predictive quantity rather than a property re-derived per configuration. We build on it a compact offset-free interaction-error MPC for torque-controlled manipulators: a force-domain random-walk state estimates persistent interaction and model error, and a 30-variable convex QP maps the correction through the current task inertia while constraining the applied joint torque. Conditional results establish impedance equivalence of the unconstrained passive feedback, offset-free regulation at feasible frozen configurations, and quadratic stabilizability of the scheduled backbone. In a 1kHz MuJoCo simulation of a 7-DOF Franka FR3, the estimator cuts steady-state error under a repeated 15N step from 2.77mm to 0.042mm when added to the otherwise identical 100Hz MPC. A stiffness-and-damping-calibrated impedance baseline attains 2.59mm but briefly saturates and needs 3.3x the peak positive joint power. Adding ideal measured-force cancellation to that baseline gives 1.39mm, so constant-load rejection is not unique to MPC; the sensorless controller still reaches 0.042mm in the moving task, a 65x reduction without force sensing and without the baseline's saturation or power cost. Demonstrated in simulation under a shared actuator budget, the contribution is an efficient operational-space realization complementing rather than replacing broader interaction-control architectures.

Yongyan Cao, Jinshan Tang

Voryx Robotics LLC, San Jose, CA 95136, USA · Department of Health Administration and Policy, George Mason University, Fairfax, VA, USA

Model Predictive ControlHuman-Robot Interaction
AskPDF
Oct 12, 2026

Adaptive Interaction Network for Human Motion Prediction During Human-Robot Collaboration

Human motion prediction during human-robot collaboration is critical for achieving safe and efficient interactions in shared environments. Unlike previous studies that have primarily focused on humans and objects, we focus on the robot-aware human motion prediction task, which explicitly models the influence of robots on human motion. This task presents new challenges, such as capturing human-robot heterogeneity and modeling complex spatial interactions. To address these issues, we develop an adaptive interaction network (AINet) model that consists of two branches: a primary branch that predicts future human motion and an auxiliary branch that estimates robot trajectories. The two branches are jointly optimized and coupled via a Local-Global Spatial Interaction (LGSI) Module, which effectively captures fine-grained and global contextual dependencies between human and robot motion sequences. Then, we introduce an Adaptive Weighted Aggregation (AWA) Module to dynamically fuse motion features using context-dependent weights, thereby increasing adaptability across diverse scenarios. Furthermore, we adopt a coarse-to-fine prediction strategy, in which a coarse pseudo-trajectory is first predicted, followed by the refinement of detailed human poses based on that trajectory. Extensive experiments on three challenging datasets reveal that our proposed method achieves superior performance, validating its effectiveness in human-robot collaboration scenarios. Our code is available at: https://github.com/LyTingHub/AINet.

Mengyuan Liu, Yangting Lin, Qiongjie Cui

State Key Laboratory of General Artificial Intelligence, Peking University, Shenzhen Graduate School, Shenzhen, Guangdong 518055, China · Guangdong Provincial Key Laboratory of Ultra High Definition Immersive Media Technology, Peking University, Shenzhen Graduate School, Shenzhen, Guangdong 518055, China · National Key Laboratory of Human-Machine Hybrid Augmented Intelligence, Xi’an Jiaotong University, Xi’an, Shaanxi 710049, China

Human Motion PredictionHuman-Robot CollaborationMulti-Agent Trajectory Prediction
AskPDF
Aug 24, 2026

Free-Energy-Gated Plasticity for Real-Time Online Motor Learning in Physical Human-Robot Interaction

Fully online embodied learning requires synaptic adaptation to acquire new behaviors while preserving previously learned dynamics during ongoing interaction. We extend the Predictive-Coding-inspired Variational Recurrent Neural Network (PV-RNN) to continuously adapt its synaptic weights and propose Free-Energy-Gated Plasticity (FEGP), which regulates the effective learning rate according to variational free energy. In real-time physical human-robot interaction, a randomly initialized network acquired three cyclic motor patterns without offline pretraining, replay, or task-boundary signals, with all three patterns emerging in autonomous rollouts. Controlled experiments over ten randomized teaching streams and five network initializations per stream showed that FEGP substantially improved repertoire coverage and retention of previously acquired patterns after they left the recent observation window. Neither a constant learning rate matched to the gate's time-averaged effective rate nor replay of the same gain values with disrupted temporal organization reproduced these improvements. These results indicate that the temporal allocation of plasticity relative to model-environment mismatch, rather than simply its average magnitude or distribution, is critical for maintaining previously acquired behaviors during continued online learning.

Hiroki Sawada, Jun Tani

Cognitive Neurorobotics Research Unit, Okinawa Institute of Science and Technology Graduate University, Okinawa, Japan 904-0302

Online Continual LearningContinual Robot LearningRecurrent Neural NetworksHuman-Robot InteractionPredictive Coding
AskPDF