cs.ROSep 25, 2026

Multi-Objective Human-in-the-Loop Bayesian Optimization of a Lower-Limb Exoskeleton

Authors: Neil Janwani, Matthew T. Lerner, Aaron J. Young, Maegan Tucker

Organizations: Georgia Institute of Technology

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Human-in-the-loop optimization (HILO) is a common approach for optimizing the control of assistive devices to account for the wearer's unique biomechanics and subjective preferences. However, despite research suggesting that a person may have a different prioritization of objectives depending on time-varying factors such as the environment, their mood, or energy levels, existing HILO approaches only consider a single objective or enforce a fixed weighting on a set of objectives. Neither approach is capable of representing an individual's preferences over objectives. In this work, we propose Multi-Objective Human-in-the-loop Bayesian Optimization (MO-HILBO), which builds on explicit multi-objective Bayesian optimization to efficiently infer a personalized set of Pareto-optimal controllers. We compare our approach with an existing multi-objective HILO method and experimentally demonstrate MO-HILBO on a lower-limb exoskeleton across two objectives: metabolic cost (efficiency) and ordinal human feedback (comfort). We find that MO-HILBO (1) discovers Pareto-optimal controllers, and (2) that the pairwise ordering of points on the Pareto front itself is consistent with validation trials. Lastly, we open-source mohilo, a Python package for running both HILO and MO-HILBO on wearable devices: https://dynamicmobility.github.io/mohilo/.

Figures & tables

Fig. 1 : Our work demonstrates MO-HILBO on a lower-limb exoskeleton (left) over two objectives: (a) metabolic cost, measured by indirect calorimetry, and (b) subjective comfort, recorded using a tablet. The protocol identifies the Pareto set of control parameters (bottom right) defined as the actions associated with the Pareto front, approximated by a learned probabilistic surrogate (top right).
Fig. 2 : MO-HILBO aims to efficiently identify the Pareto set of mutually-optimal actions across multiple objectives. Similar to traditional HILO, MO-HILBO cycles over three main steps: 1) collecting feedback from human users, 2) updating the belief of the underlying objective function(s) based on the provided feedback, and 3) selecting new queries for the subsequent iteration using an acquisition function. In our work, we explore the application of MO-HILBO on a lower-limb exoskeleton across two objectives: subjective evaluations of comfort using ordinal labels, and metabolic cost measured by indirect calorimetry.
Fig. 3 : Synthetic experiments on high noise and limited sample budget. Across the ablation studies (left three plots) MO-HILBO achieves a lower IGD+\mathrm{IGD}^{+} score across different noise levels, action dimensions, and numbers of objectives than NSGA-II. All curves are averaged over 5050 runs to better measure performance.
Fig. 4 : Subject MB03’s Pareto set progression throughout the optimization procedure. The Pareto set begins as the entire action space, and eventually clusters around low delay, low hip extension scale. For MB03, hip flexion scale was the primary decision variable for comfort versus efficiency. In fact, MB03’s most comfortable gait was when the exoskeleton was turned off, while their most efficient controller was when the hip flexion torque scale was nearly maximized. In short, by performing MO-HILBO, we are able to reduce the search space of controllers from the entire action box, visualized on the bottom left, to a simpler, single-dimension Pareto set, on the bottom right.
Fig. 5 : Predicted versus measured points (a), (b) and discovered Pareto sets (c). (a) shows a fitted MO-GP on MB02, where the training data lie near the Pareto front–a result of MO-HILBO’s acquisition function exploring and refining the Pareto set. (b) shows the validation points (stars)–which were not seen during training–spaced out over the front. (c) shows the overlayed Pareto sets of all subjects. For the vast majority of points, each subject’s Pareto set does not overlap in action space, except for part of MB03’s and MB01’s. While the subjects’ Pareto-optimal controllers featured a variety of flexion or extension torques, none had a high delay.
 HypervolumePairwise Ordering
SubjectParetoAnti-ParetoMetabolic CostComfort
MB014.021.300100% (3/3)100% (3/3)
MB021.550.366100% (3/3)100% (3/3)
MB031.020.12867% (2/3)100% (3/3)
TABLE I : Validation data across subjects. In all cases, the identified Pareto sets were validated to Pareto-dominate the anti-Pareto set (built by flipping the sign of each objective). MO-HILBO also correctly predicted 94% (17/18) of the pairwise orderings of Pareto-optimal points. A valid reference point was chosen for each subject’s hypervolume computation, resulting in different scales.
More in this topic

Human-in-the-Loop AI

Human-in-the-loop AI combines human feedback, oversight, and decision-making with machine learning and autonomous systems.

See all 288 papers

Explore similar work

CardsList
Aug 10, 2026

Personalized Lower-limb Exoskeleton Assistance via Preference-based Bayesian Optimization

A significant challenge in exoskeleton robotics is the need to dynamically adapt control profiles to individual motion preferences, thereby ensuring both efficient and comfortable assistance. Currently, since user experience can serve as a comprehensive metric for evaluating the effectiveness of assistance, user preference-based optimization methods have been widely studied for parameter tuning. However, the existing methods rely heavily on extensive human-robot online interactions and suffer from slow optimization speed, which not only induces user fatigue but also compromises optimization effectiveness. Therefore, this paper aims to explore an efficient preference-based optimization framework for personalized exoskeleton assistance that can learn optimal parameters with minimal interaction. We propose a preference-based Bayesian optimization (PbBO) approach that can improve sample efficiency by leveraging knowledge about the sampling distribution of candidate sets. For optimizing six control parameters, PbBO can converge to user-preferred parameters with 90.7% validation accuracy via 20 iterations. Moreover, the hierarchical controller is designed to generate personalized torque for different tasks and achieve interaction torque tracking in real time. The results of treadmill and outdoor experiments demonstrate that the optimized parameters can reduce metabolic rate by 14.5%-15.4%, heart rate by 6.3%-7.6%, and muscle activation by 6.7%-31.5% compared to unassisted walking.

Xiao-Yin Liu, Guotao Li, Weiqun Wang +1

State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China · School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China · CASIA-MUST Joint Laboratory of Intelligence Science and Technology, Institute of Systems Engineering, Macau University of Science and Technology, Macao, China

Bayesian OptimizationAnkle Exoskeleton ControlAssistive Robotics
AskPDF
Feb 26, 2026

Bayesian Preference Elicitation: Human-In-The-Loop Optimization of An Active Prosthesis

Tuning active prostheses for people with amputation is time-consuming and relies on metrics that may not fully reflect user needs. We introduce a human-in-the-loop optimization (HILO) approach that leverages direct user preferences to personalize a standard four-parameter prosthesis controller efficiently. Our method employs preference-based Multiobjective Bayesian Optimization that uses a state-or-the-art acquisition function especially designed for preference learning, and includes two algorithmic variants: a discrete version (EUBO-LineCoSpar), and a continuous version (BPE4Prost). Simulation results on benchmark functions and real-application trials demonstrate efficient convergence, robust preference elicitation, and measurable biomechanical improvements, illustrating the potential of preference-driven tuning for user-centered prosthesis control.

Sophia Taddei, Wouter Koppen, Eligia Alfio +8

Federated Labs Robotics & AI, Vrije Universiteit Brussel. · Human Physiology and Sports Physiotherapy Research Group, Vrije Universiteit Brussel. · IDLab, Ghent University – imec. +1

Preference ElicitationHuman-in-the-Loop AIBayesian Optimization
AskPDF
Sep 18, 2026

Simultaneous Forward and Inverse Human-in-the-Loop Optimization

Subjective user experience is important to human-robot interaction, but the outcomes users value, and how those preferences vary across individuals and contexts, are often unknown. While inverse learning approaches using human data can help identify user rewards, in many assistive settings the experimental costs of executing a control policy, measuring biomechanical or physiological outcomes, and collecting user feedback often limit the number of queries and optimization iterations. Here, we present Simultaneous Forward and Inverse Human-In-the-Loop Optimization (SFIHILO), which efficiently infers individual-specific reward functions from preferences over human outcomes and identifies a final control policy that maximizes the learned reward. SFIHILO bootstraps a forward model to predict user outcomes from control policies, uses this model for active querying to accelerate inverse reward learning, and then optimizes the final control policy without additional user trials. We score candidate policies by their expected reduction in uncertainty across both forward and inverse beliefs, targeting a regional preference boundary to robustly inform this simultaneous learning process. In simulation, we show that SFIHILO was effective across user heterogeneity, outcome dimensionalities, precision requirements, noise levels, and nonstationarity; compared with mutual information approaches, the proposed active querying strategy significantly improved sample efficiency in inverse learning while preserving forward model accuracy. This approach demonstrates the potential to infer latent human goals, enabling more transferable and effective human-robot interaction.

Kyeongwon Park, Steven H. Collins

Department of Mechanical Engineering, Stanford University, United States

Human-in-the-Loop AIHuman-Robot InteractionPreference-Based RLAssistive RoboticsReinforcement Learning
AskPDF