Flexible Shape-Sensing Breakthrough Enables 3D Perception for Soft Robotics
A newly developed double-layer orthogonal fiber Bragg grating (FBG) sensor has achieved high-precision three-dimensional shape reconstruction for soft robotic systems, marking a significant advancement in flexible tactile feedback technologies. This innovation, led by Yongxing Guo and colleagues at Wuhan University of Science and Technology, addresses persistent challenges in curvature sensing, spatial coordinate mapping, and surface contour fidelity—critical gaps that have hindered the deployment of soft robots in complex, unstructured environments.
The sensor leverages polydimethylsiloxane (PDMS) as a flexible substrate, into which 32 FBGs are embedded in an orthogonal double-layer array. Unlike prior approaches that only enabled one-dimensional curvature tracking along a single fiber axis, this architecture captures spatial deformation across both horizontal and vertical planes simultaneously. By converting wavelength shifts into curvature data and then into 3D coordinates through interpolation and fitting algorithms, the system reconstructs the full surface geometry of objects with a measurement error ranging from just 2.8% to 4.5%.
This development is particularly timely as soft robotics enters a new phase of real-world application. From minimally invasive surgical tools to wearable assistive devices and adaptive grippers in logistics automation, soft robots offer unparalleled compliance and safety when interacting with humans or delicate objects. However, their very flexibility—while advantageous—has posed a fundamental sensing dilemma: how to measure shape without compromising compliance. Traditional strain gauges or electromagnetic sensors are too rigid, bulky, or susceptible to electromagnetic interference, making them incompatible with the elastomeric materials that define soft robotics.
Fiber Bragg gratings have long been recognized as a promising alternative due to their inherent flexibility, immunity to electromagnetic noise, and capacity for multiplexed measurements along a single optical fiber. Yet most existing FBG-based shape sensors remain limited to linear or quasi-2D reconstructions. They typically assume pure bending along one axis or rely on rigid polymer substrates like polyvinyl chloride (PVC) or polyimide, which restrict full deformation adaptability. The new sensor overcomes these limitations by using PDMS—a biocompatible, optically transparent silicone elastomer—with a carefully optimized curing ratio (1:5 base to crosslinker) that balances mechanical softness and stable strain transfer.
In experimental validation, the team fabricated a sensor prototype measuring 80 mm by 80 mm by 3 mm, with two layers of four optical fibers each. Each fiber hosted four equally spaced FBGs, yielding a 4×4 sensing grid per layer. The upper-layer fiber cores were positioned 1.125 mm from the neutral bending plane, while the lower-layer cores sat 0.625 mm away, enabling differential strain detection during bidirectional bending. Calibration tests under controlled temperature (25°C) confirmed a linear relationship between curvature and wavelength shift, with sensitivities of 57.641 pm/m⁻¹ for the upper array and 32.250 pm/m⁻¹ for the lower array—demonstrating both high responsiveness and directional discrimination.
Crucially, the design exhibits near-zero cross-talk: when the sensor was bent to activate only the upper FBG layer, the lower layer showed negligible wavelength drift, and vice versa. This confirms the system’s ability to isolate orthogonal curvature components, a prerequisite for accurate 3D reconstruction.
To test real-world performance, the researchers 3D-printed three distinct test surfaces: a constant-curvature dome, a sinusoidal wave pattern, and a mouse-shaped contour mimicking ergonomic object geometry. The sensor was placed over each surface and held in conformal contact for over 30 minutes to ensure thermal and mechanical stabilization. The resulting wavelength data from all 32 FBGs were processed through a coordinate transformation framework derived from material mechanics and circular arc geometry. Using MATLAB-based interpolation, the team reconstructed continuous 3D surface models that closely matched the original test geometries.
While the reconstructions captured the global shape accurately, minor deviations appeared between discrete sensing points—highlighting a known limitation of interpolation-based methods. The authors acknowledge three primary error sources: (1) the assumption of uniform curvature across each FBG’s physical length, which may not hold under complex deformations; (2) residual thermal fluctuations despite ambient control, potentially induced by human handling; and (3) microscale misalignment during manual FBG embedding. These are systematic rather than fundamental issues, suggesting clear paths for future refinement through algorithmic optimization and automated fabrication.
The implications extend beyond soft robotics. This sensing modality could enhance haptic feedback in virtual reality gloves, enable shape monitoring in inflatable space structures, or support real-time deformation tracking in prosthetic limbs. Its all-optical nature also ensures compatibility with MRI environments and explosive atmospheres where electronic sensors fail.
From an engineering perspective, the work stands out for its integration of materials science, optical physics, and computational geometry. The PDMS substrate not only provides mechanical compliance but also allows visual inspection of embedded FBGs during assembly—a practical advantage often overlooked in academic prototypes. The use of a custom-built wavelength demodulator with 3-pm resolution further underscores the team’s commitment to measurement rigor.
Moreover, the approach aligns with emerging standards for trustworthy AI-augmented sensing systems. By grounding reconstruction in first-principles physics rather than black-box machine learning, the method offers interpretability and traceability—key components of Google’s EEAT (Experience, Expertise, Authoritativeness, Trustworthiness) framework. The linear calibration curves and quantified error margins provide transparency that regulators and end-users increasingly demand.
Looking ahead, the research group plans to explore higher-density FBG arrays, integration with active actuation for closed-loop control, and miniaturization for endoscopic applications. They also aim to replace manual fiber placement with microfluidic-assisted embedding to reduce fabrication variability. Such improvements could push curvature measurement errors below 2%, approaching the precision of rigid industrial metrology tools—while retaining full flexibility.
This work exemplifies how interdisciplinary convergence can solve persistent bottlenecks in emerging technologies. By rethinking not just the sensor architecture but also the substrate chemistry and reconstruction mathematics, Guo and his team have delivered a platform that bridges the gap between theoretical shape sensing and practical soft robotic deployment.
Yongxing Guo, Yuehui Yang, Li Xiong
Key Laboratory of Metallurgical Equipment and Control Technology, Ministry of Education, Wuhan University of Science and Technology, Wuhan 430081, China
Hubei Key Laboratory of Mechanical Transmission and Manufacturing Engineering, Wuhan University of Science and Technology, Wuhan 430081, China
Published in Optics and Precision Engineering, Vol. 29, No. 10, October 2021
DOI: 10.37188/OPE.20212910.2306