💻 computer science

Simulation-based multi-criteria comparison of mono-articular and bi-articular exoskeletons during walking with and without load

This paper presents a simulation-based multi-criteria Pareto optimization study comparing mono-articular and bi-articular exoskeletons, revealing that while bi-articular designs are less sensitive to load regarding power consumption, mono-articular configurations better reduce peak joint reaction forces, with device inertia having minimal impact on the metabolic cost of bi-articular solutions.

Original authors: Ali KhalilianMotamed Bonab, Volkan Patoglu

Published 2026-05-14

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Ali KhalilianMotamed Bonab, Volkan Patoglu

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine your legs are a complex construction site. You have muscles acting as workers, joints acting as hinges, and your brain acting as the foreman, trying to get you from point A to point B using the least amount of energy possible. Sometimes, carrying a heavy backpack or getting older makes this job much harder, and your "workers" (muscles) get tired faster.

This paper is about designing robotic suits (exoskeletons) that act like helpful assistants to these workers. The researchers wanted to figure out which type of assistant is better: one that helps each joint individually, or one that helps two joints at once using a clever mechanical trick.

Here is a breakdown of their study using simple analogies:

1. The Two Types of Assistants

The researchers compared two designs for a robotic suit that helps the hip and knee:

The "Single-Tasker" (Mono-articular): Imagine a worker who has a separate motor attached directly to your hip and another separate motor attached directly to your knee. Each motor only knows how to push or pull its specific joint. This is simple and direct, like having two separate electric fans.
The "Team Player" (Bi-articular): Imagine a worker who has one big motor on your hip, but it uses a system of ropes and pulleys (a parallelogram mechanism) to send power down to your knee. This mimics how human muscles (like the hamstrings) work: they cross over two joints. This design keeps the heavy motor near your body (the hip) rather than on your lower leg, making the leg feel lighter and easier to swing.

2. The Virtual Test Lab

Testing these robots on real people is expensive, slow, and risky. You'd have to build many prototypes, find volunteers, and worry about safety.

Instead, the researchers built a digital twin of a human body inside a computer. They simulated 7 different people walking, both with and without a heavy 38kg (84 lb) backpack. They didn't just watch them walk; they calculated exactly how much energy the "muscles" used and how much power the "robot motors" consumed.

3. The Great Balancing Act (The Trade-off)

The researchers faced a classic dilemma: Help more vs. Use less battery.

If the robot pushes really hard to save your energy, it needs a big battery and uses a lot of electricity.
If the robot uses very little electricity, it might not help you much.

To solve this, they used a method called Pareto Optimization. Think of this as finding the "best possible deals" on a menu. You can't have the biggest burger and the cheapest price at the same time, but there is a specific size and price point where you get the most satisfaction for your money. They found the "sweet spot" designs where the robot gives the maximum help for the minimum battery drain.

4. What They Discovered

The "Ideal" Scenario (No Limits)
First, they imagined a perfect robot with infinite power and no weight. In this fantasy world, both the "Single-Tasker" and the "Team Player" were equally good at saving energy. They could both reduce the energy cost of walking by about 22% when walking without a load, and about 20% when carrying a heavy load.

The "Real World" Scenario (With Limits)
Then, they added real-world limits: the motors can only push so hard (torque limits), and the robot has weight.

The Heavy Load Effect: When people carried a heavy backpack, the "Single-Tasker" had to work much harder and use more battery power to keep up. The "Team Player" was more stable; its power usage didn't change as drastically when the load increased.
The Weight Problem: Robots have weight, and carrying extra weight on your legs is tiring. The "Team Player" kept its heavy motor near the hip (close to the body), while the "Single-Tasker" had motors on the lower leg. The study found that the "Team Player" was much better at handling its own weight. The "Single-Tasker" lost a lot of its efficiency because the extra weight on the lower leg made walking harder.
The "Regeneration" Trick: Just like a hybrid car brakes to recharge its battery, these robots can capture energy when they slow down (negative power). The study found that the "Single-Tasker" had a lot of energy to capture (especially at the knee), while the "Team Player" was already efficient enough that it didn't need to capture as much to be effective.

The Surprising Winner
Despite the "Single-Tasker" being simpler to build, the "Team Player" (Bi-articular) came out as the superior design for most scenarios.

It was less sensitive to heavy loads.
It handled its own weight much better, preserving the user's energy.
It provided a more predictable and consistent helping pattern, regardless of whether the person was carrying a backpack or not.

5. The "Hidden" Benefits

The study also looked at what happens inside the body. They found that when the robot helps, it doesn't just change how the hip and knee move; it changes how the ankle and hip muscles work too.

Joint Protection: The robots significantly reduced the "crushing force" (reaction forces) inside the knee and hip joints. Imagine the robot acting like a shock absorber, taking the brunt of the impact so your joints don't have to. This is good for preventing wear and tear.

Summary

The researchers used a computer simulation to prove that a robotic suit designed like a human muscle (crossing two joints) is generally a smarter design than one with motors on every joint. It saves more energy, handles heavy loads better, and is less taxing on the body because it keeps the heavy parts of the robot closer to the body's center.

Note: The paper strictly limits its claims to these simulation results. It does not claim these robots are ready for hospitals or daily use yet; it simply provides a "blueprint" and guidelines for engineers on how to build better ones in the future.

Technical Summary: Simulation-Based Multi-Criteria Comparison of Mono-articular and Bi-articular Exoskeletons

Problem Statement

Developing wearable robotic assistive devices to reduce the metabolic cost of human locomotion is a significant challenge. A systematic design approach is required to capture the complex interactions between device dynamics, assistance torque profiles, and human performance. Conducting rigorous comparisons across a wide range of exoskeleton designs and torque profiles through physical human-in-the-loop experiments is generally infeasible due to the high costs of prototyping, the difficulty of recruiting sufficient volunteers, safety concerns, and the inability to measure internal physiological data (e.g., muscle fiber mechanics) without invasive sensors. Furthermore, existing studies often lack a systematic mechatronic design framework that accounts for realistic physical constraints, such as actuator torque saturation, power consumption, and device inertia.

Methodology

The authors propose a simulation-based multi-criteria design approach using the OpenSim framework to systematically evaluate and compare mono-articular and bi-articular hip-knee exoskeletons.

1. Musculoskeletal Modeling and Simulation:

Subjects and Conditions: Simulations were conducted on seven subjects walking at self-selected speeds under two conditions: no load and carrying a 38 kg torso load.
Model: A 39-degree-of-freedom, 80-muscle musculoskeletal model (based on Rajagopal et al.) was scaled to each subject.
Workflow: The study utilized a standard OpenSim workflow: scaling, inverse kinematics, residual reduction algorithm (RRA), and computed muscle control (CMC).
Assistive Device Integration: The CMC algorithm was modified to include assistive torques from "torque actuators" as part of the objective function. This allowed the optimizer to utilize exoskeleton assistance to track kinematics while minimizing muscle activation and reserve actuator usage.
Device Configurations:
- Mono-articular: Modeled as a two-link serial manipulator with actuators directly mounted at the hip and knee joints.
- Bi-articular: Inspired by human bi-articular muscles, this design places all actuators at the hip (proximal) and delivers power to the knee (distal) via a parallelogram mechanism. This aims to reduce distal leg inertia and enable power transformation between joints.

2. Multi-Criteria Optimization (Pareto Optimization):

Objectives: The study simultaneously optimized two conflicting objectives: maximizing metabolic cost reduction and minimizing device power consumption.
Constraints: Realistic physical limits were introduced by systematically constraining the peak torque of the actuators (ranging from 30 Nm to 70 Nm). This introduced actuator saturation, breaking the linear kinematic mapping between the two device types.
Method: An $\epsilon$ -constraint method was used to generate Pareto-front curves, representing the set of non-dominated solutions where improving one objective degrades the other.

3. Secondary Analyses:

Regeneration: The potential for energy regeneration (harvesting negative work) was superimposed on the Pareto solutions with efficiencies up to 65%.
Inertia: The detrimental metabolic effects of added device mass and inertia were estimated using a model by Browning et al. and superimposed on the Pareto fronts. A modified augmentation factor (MAF) was introduced to quantify the net metabolic benefit considering these inertial penalties.
Joint Loads: Joint reaction forces and moments were computed to assess the impact of assistance on joint health (specifically knee, patellofemoral, and hip joints).

Key Contributions

Systematic Design Framework: Introduction of a Pareto optimization approach to rigorously compare exoskeleton designs under realistic constraints (torque limits, power consumption, and inertia), moving beyond "ideal" unconstrained simulations.
Bi-articular vs. Mono-articular Comparison: A direct, fair comparison of two distinct kinematic architectures. The study demonstrates that while ideal devices (unconstrained) perform similarly, their behavior diverges significantly under actuator saturation and inertial constraints.
Inertial and Regeneration Effects: Quantification of how device inertia and power regeneration capabilities alter the optimal design space. The study introduces a modified augmentation factor to better evaluate net metabolic efficiency.
Load Adaptation: Analysis of how heavy loads (38 kg) affect optimal assistance profiles, muscle coordination, and device performance.

Key Results

1. Ideal vs. Torque-Limited Performance:

Under ideal conditions (no torque limits), both mono-articular and bi-articular exoskeletons achieved similar metabolic cost reductions (~22% for no load, ~20% for loaded).
Under torque-limited conditions, both configurations could still achieve near-ideal metabolic reductions, but with significantly lower power consumption. However, the mono-articular design required larger peak torque limits to achieve comparable performance to the bi-articular design.

2. Power Consumption and Sensitivity to Load:

Mono-articular: Power consumption was more sensitive to loading conditions. The power distribution shifted significantly, with hip actuators becoming dominant under load. The torque and power profiles of non-dominated mono-articular solutions showed high variation and were less predictable across load conditions.
Bi-articular: Power consumption was less sensitive to loading. The power distribution between hip and knee actuators remained relatively uniform regardless of the load. The torque profiles were more consistent and predictable between no-load and loaded conditions.

3. Joint Reaction Forces:

Both device types significantly reduced peak reaction forces at the knee, patellofemoral, and hip joints.
Mono-articular: Demonstrated superior performance in reducing peak reaction forces and moments compared to the bi-articular design, particularly at the knee and patellofemoral joints.
Bi-articular: Showed lower peak reaction moments in the extension-flexion direction compared to mono-articular devices.

4. Regeneration and Inertia:

Regeneration: Including regeneration (up to 65% efficiency) improved power consumption significantly (6.54% to 25.76%). Mono-articular knee actuators showed higher regeneration potential than hip actuators, whereas bi-articular actuators showed similar potential across both joints.
Inertia: Added inertia significantly degraded metabolic efficiency. Mono-articular devices lost 42.51%–55.51% of their efficiency due to inertia, while bi-articular devices lost 35.12%–49.67%.
Pareto Shift: Inertial effects shifted the Pareto front for mono-articular devices, changing the set of non-dominated solutions. In contrast, the bi-articular Pareto-optimal configurations remained largely unchanged, though their overall efficiency was reduced.

5. Muscle Coordination:

Assistance torques indirectly affected muscles not directly assisted (e.g., ankle plantarflexors and hip abductors).
Loading caused predictable scaling and time-shifts in torque profiles and muscle activations.
The bi-articular design's ability to keep actuators proximal reduced the reflected inertia, aligning with human anatomical efficiency.

Significance and Claims

The paper claims that its simulation-based multi-criteria approach provides a rigorous and fair methodology for comparing exoskeleton designs, overcoming the limitations of experimental-only studies.

Design Guidelines: The results suggest that bi-articular exoskeletons are advantageous for their favorable inertial distribution and robustness to varying load conditions, making them potentially easier to control with generic strategies across different tasks.
Trade-offs: While mono-articular exoskeletons may offer better reduction in peak joint reaction forces (potentially beneficial for joint health), they are more sensitive to inertial penalties and require higher torque capacities to match the metabolic benefits of bi-articular designs.
Realism: The study emphasizes that ignoring actuator saturation and inertial effects leads to misleading design conclusions. The proposed framework allows designers to visualize the trade-offs between metabolic savings, power consumption, and device weight/inertia before physical prototyping.
Future Validation: The authors state that these simulation results serve as design guidelines to identify promising configurations for future human-in-the-loop experimental validation, rather than claiming immediate clinical efficacy.

The study concludes that while bi-articular designs offer superior performance in terms of metabolic efficiency and load insensitivity, mono-articular designs may still be viable if inertial properties are carefully managed (e.g., by placing knee actuators more proximally) and if the primary goal is the reduction of peak joint reaction forces.