Multi-objective optimization-based design of a compliant gravity balancing orthosis: development and validation

This paper presents a multi-objective optimization framework using particle swarm optimization to design a compliant, 3D-printed shoulder orthosis that effectively reduces muscle activity in healthy individuals while balancing gravity compensation with wearability constraints.

The Gist

Imagine your shoulder is like a heavy door on a hinge. When you lift your arm, gravity tries to slam that door shut. For people with weak muscles or injuries, holding that door open is exhausting, like trying to keep a heavy car door propped open with just your pinky finger.

This paper is about building a smart, stretchy "doorstop" (an orthosis) that automatically holds the door open for you, so your muscles can take a break. But instead of using heavy springs or motors, the researchers used a special kind of flexible plastic that bends and twists to do the work.

Here is the story of how they built it, explained simply:

1. The Problem: It's Not Just About Strength

Previous devices were good at holding the arm up, but they were often bulky, uncomfortable, or pushed on your skin in weird places. It's like wearing a backpack that holds your arm up but also digs into your ribs and feels like a brick.

The researchers wanted to design a device that:

• Holds the arm up perfectly (like a perfect counterweight).
• Is comfortable (doesn't poke you).
• Is small enough to wear without looking like a robot.

2. The Solution: A "Digital Architect"

Since designing a piece of plastic that bends just right is incredibly hard to do by hand, they built a computer brain to do the thinking for them.

Think of this computer brain as a super-taster at a restaurant.

• The Menu: The computer generates thousands of different "recipes" (shapes) for the plastic brace.
• The Taste Test: It simulates wearing the brace on a virtual human.
• The Critique: It checks two things:
  1. Does it hold the arm up well? (The "Flavor" score).
  2. Does it hurt or look weird? (The "Comfort" score).

The computer uses a method called Particle Swarm Optimization. Imagine a flock of birds searching for the best berry bush. Each bird (a potential design) flies around, checks the berries, and tells the others where the best spots are. Over time, the whole flock converges on the absolute best spot.

3. The Four Rounds of "Try, Try Again"

The researchers didn't just run the computer once; they ran it in four increasingly difficult rounds to make sure the design was practical:

• Round 1 (The Dream): The computer designed a perfect shape that held the arm up perfectly. But, when they checked, the shape was huge and would crash into the person's body. It was like a perfect umbrella that was too big to fit in a car.
• Round 2 (The Reality Check): They told the computer, "No crashing into the body!" The computer redesigned the brace to wrap around the person. It wasn't quite as perfect at holding the arm up, but it actually fit on a human.
• Round 3 (The Balancing Act): They added a new rule: "Don't push too hard on the wrist." If the brace pushes too hard on the end of the arm, it feels like someone is poking you. The computer had to find a "Goldilocks" zone: strong enough to hold the arm, but gentle enough not to poke.
• Round 4 (The Real World): Finally, they told the computer, "Make sure this fits on a 3D printer and uses real-world measurements." This gave them the final, printable design.

4. The Human Test: Does it Work?

They took the best designs, printed them out using a flexible, rubbery plastic (like a high-tech gummy bear), and put them on 6 healthy volunteers.

The Results were amazing:

• The "Front Shoulder" Muscle: When wearing the brace, this muscle relaxed by 53%. It's like taking a heavy backpack off your shoulders and letting your muscles rest.
• The "Upper Back" Muscle: This muscle relaxed by a massive 71%.
• The "Back Shoulder" Muscle: Even the muscle that usually fights against the brace relaxed, suggesting the brace was so smooth it didn't confuse the body.

However, there was a small catch: The chest muscle worked a little harder. Why? Because the brace was designed in a 2D world (flat), but humans are 3D. As the brace bent, it pushed the arm slightly sideways, and the chest muscle had to work to keep the arm from swinging out.

The Big Takeaway

This paper proves that we can use AI and math to design wearable medical devices that aren't just functional, but also comfortable and human-friendly.

Instead of guessing what shape a brace should be, we can let a computer "evolve" thousands of designs to find the perfect one that balances strength and comfort. It's a huge step toward making wearable tech that feels less like a machine and more like a helpful friend.

Technical Summary

1. Problem Statement

Shoulder disabilities and muscle weakness (e.g., from stroke or neurological disorders) severely limit the ability to perform Activities of Daily Living (ADLs), particularly arm elevation. While passive elastic exoskeletons can assist by providing gravity compensation, existing designs often prioritize mechanical performance (gravity balancing) at the expense of wearability.

• The Gap: Current optimization frameworks for compliant mechanisms focus almost exclusively on achieving the ideal compensation moment profile. They often neglect critical wearability constraints such as device size, bulk, and the generation of uncomfortable concentrated loads at attachment points.
• The Goal: To develop a design framework that balances high-fidelity gravity compensation with practical wearability constraints (size, comfort, and avoidance of undesirable loading directions) for a soft, compliant shoulder orthosis.

2. Methodology

A. Optimization Framework

The authors developed a Multi-Objective Optimization (MOO) framework integrating Finite Element Analysis (FEA) with Particle Swarm Optimization (PSO).

• Simulation: 2D nonlinear static simulations were performed in SolidWorks to model the large deformations of the compliant mechanism.
• Optimization: A Particle Swarm Optimizer (PSO) implemented in MATLAB guided the search for optimal designs. A custom VB.NET application automated data transfer between the simulation and optimizer.
• Design Encoding: The orthosis was modeled as a monolithic structure made of Thermoplastic Polyurethane (TPU). Its shape was encoded via a design vector defining:
  • Attachment points (proximal on torso, distal on arm).
  • A linkage chain defining the neutral axis (B-spline).
  • In-plane thickness profiles.
  • Out-of-plane width.
• Objectives:
  1. Gravity Compensation ($\hat{f}_G$): Minimizing the error between the generated moment and the ideal sinusoidal moment required to balance the arm ($M = mgL \sin \theta$).
  2. Distal Moment Minimization ($\hat{f}_M$): Minimizing concentrated moments at the distal attachment point to prevent user discomfort.
  3. Constraints: Nonlinear constraints were enforced to prevent collisions (undeformed/deformed states), ensure full rotation, and adhere to 3D printer bed limits.

B. Optimization Sequence

The study employed a staged approach to progressively increase complexity:

1. Stage 1 (Single-Objective, Out-of-Plane, Adimensional): Validated if sinusoidal profiles were achievable without collision constraints. Resulted in high accuracy but impractical, oversized designs.
2. Stage 2 (Single-Objective, In-Plane, Adimensional): Added collision constraints. Accuracy dropped slightly (4.2% error), but designs became wearable.
3. Stage 3 (Multi-Objective, In-Plane, Adimensional): Introduced the trade-off between accuracy and distal moment. Generated a Pareto front of solutions.
4. Stage 4 (Multi-Objective, In-Plane, Dimensional): Added physical dimensions (target moment = 8 Nm for a 50th percentile male) and printer bounds.

C. Prototyping and Validation

• Fabrication: Two designs were selected from the Pareto fronts (Design E from Stage 3 and Design F from Stage 4) and 3D printed using TPU.
• In-Vivo Study: 6 healthy participants performed static and dynamic shoulder flexion tasks (with and without a 5 lb weight).
• Measurements: Surface Electromyography (sEMG) recorded activity in the Anterior Deltoid (AD), Pectoralis Major (PM), Upper Trapezius (UT), and Posterior Deltoid (PD).
• Analysis: Linear mixed models were used to determine statistical significance of muscle activity reduction.

3. Key Contributions

1. Novel Optimization Framework: A custom MOO framework that integrates FEA and PSO to simultaneously optimize for mechanical performance and wearability constraints (size, collisions, and comfort).
2. Staged Design Strategy: A systematic progression from simplified adimensional problems to fully dimensional, constrained designs, revealing how constraints impact solution spaces.
3. Wearable Compliant Orthosis: The successful realization of a soft, 3D-printed orthosis that achieves significant muscle activity reduction without the bulk of rigid mechanisms.
4. Empirical Validation: Comprehensive in-vivo testing demonstrating the physiological efficacy of the design across static and dynamic tasks.

4. Results

Optimization Outcomes

• Accuracy vs. Wearability Trade-off: Stage 1 achieved 0.7% error but was unwearable. Stage 2 (with collisions) had 4.2% error. Stage 3 (Multi-Objective) produced a Pareto front where Design E offered a balance (11.1% error, 20.9% distal moment), while Design F (Stage 4) offered the best dimensional accuracy (3.6% error) but higher distal moment.
• Design Evolution: As the optimization prioritized minimizing distal moments, the optimal shape shifted from a "gentle, two-corner" design to a "single, harsher bend."

In-Vivo Performance

• Muscle Activity Reduction:
  • Anterior Deltoid (AD): Significant reduction of 53% in functional postures (no weight).
  • Upper Trapezius (UT): Significant reduction of 71% in functional postures (no weight).
  • Posterior Deltoid (PD): Significant reductions observed (up to 47%), suggesting the orthosis reduces the need for the antagonist muscle to stabilize the joint.
• Dynamic Tasks: Reductions were observed in dynamic tasks but were generally smaller in magnitude than static tasks, likely due to the inertial forces not being fully compensated by the passive system.
• Side Effects: A significant increase (37%) was observed in the Pectoralis Major (PM) during weighted conditions. The authors attribute this to lateral deflection of the 2D-designed orthosis, requiring the user to engage the PM to stabilize lateral motion.

5. Significance and Conclusion

This work demonstrates that multi-objective optimization is essential for designing practical compliant orthoses. By explicitly optimizing for wearability constraints (like distal loading and size) alongside gravity compensation, the authors successfully created a device that significantly reduces muscle effort (up to 71%) for users.

Implications:

• Rehabilitation & Occupational Safety: The device offers a lightweight, passive solution for stroke rehabilitation and occupational shoulder support.
• Design Paradigm Shift: The study highlights that purely performance-driven designs often fail in real-world application due to comfort and size issues. The proposed framework provides a pathway to navigate these trade-offs.
• Future Work: The authors note the need to incorporate out-of-plane load simulations to address the lateral instability observed in the PM muscle and to expand the optimization to include a third objective for device size.