Page 19 - Read Online

Page 16 of 26 Chen et al. Soft Sci. 2026, 6, 3

prolonged biostability or predictable non-toxic biodegradability [199,200] . Biostable materials must preserve their
structural and functional integrity under physiological conditions, including mucus-lined surfaces, variable
pH, enzymatic environments, and dynamic shear flows. Biodegradability necessitates advanced chemical
strategies for stable matrix-filler integration, alongside predictive degradation kinetics to ensure device safety
over the intended functional lifetime. Beyond chemical strategies, future approaches are exploring the
integration of patient-derived substances with magnetically actuatable materials to mitigate foreign-body
responses while preserving functional performance. Gel fibers prepared from magnetic fillers and the
patient's own blood have demonstrated immune evasion, controllable locomotion in physiological fluids, and
on-demand drug release in large animal models . Such advances suggest that personalized biohybrid
[94]
materials may expand the translational potential of magnetic soft robots by enabling safer clearance pathways
and facilitating targeted interventions in sensitive clinical scenarios. In addition, advanced surface
engineering of the magnetic soft robot, such as the hydrogel layer and adhesion layer, can enhance retention
and reduce mechanical irritation [201,202] . Ultimately, comprehensive in vivo studies over chronic durations,
preferably in large animal models, are essential to evaluate immune response, tissue remodeling, and
clearance or retention properties. In this context, it is also important to recognize that post-use retrieval is
feasible for robots engineered from non-degradable materials. Millimeter-scale wireless non-degradable
robots can be reliably retrieved after use. Depending on the target organ, they may be extracted using
standard minimally invasive clinical tools such as endoscopes or ureteroscopes . This enables controlled
[12]
[188]
removal and prevents long-term retention, complementing material-driven strategies for safe clinical
deployment.

Adaptive control in vivo
To achieve clinical utility, robots must operate autonomously in dynamic and unpredictable physiological
environments [50,203] . This necessitates a shift from operator-controlled open-loop systems to intelligent,
sensor-integrated platforms for real-time environmental feedback. Soft embedded sensors that detect
pressure, strain, temperature, or biochemical markers can provide robots with physiological awareness.
Recent progress in conductive soft materials for flexible electronics has further expanded the feasibility of
closed-loop magnetic soft robots [204-206] . Highly compliant and stretchable conductive polymers and
hydrogels, such as those based on poly(3,4-ethylenedioxythiophene):polystyrene sulfonate
(PEDOT:PSS) [207-210] , offer lightweight structures with excellent electrical properties and are well-suited for
incorporation into small magnetic soft robots without compromising mechanical compliance. In addition,
the combination of embedded ultrasonic soft sensors with magnetic actuators has resulted in fully wireless
microrobotic systems capable of feedback control, precise drug release, and real-time physiological
monitoring [211] . These capabilities have already been demonstrated in rabbit and porcine models, providing
concrete evidence that sensor-integrated magnetic soft robots are moving toward intelligent operation in
clinically relevant environments.

When magnetic soft robots are required to perform dynamic tasks or navigate unstructured or highly
constrained environments, learning-based control strategies may provide advantages over classical model-
based approaches, particularly when training datasets can be efficiently collected from physical experiments.
Recent work has shown that probabilistic learning frameworks can optimize gait patterns of wireless
miniature magnetic soft robots using relatively small experimental datasets [201] . Although current progress
remains limited, learning-driven control represents a promising avenue for enabling higher degrees of
autonomy and adaptability. Moreover, effective tracking of magnetic soft robots in complex biological
environments remains a significant challenge. Biological tissues, composed of heterogeneous and deformable
structures, may shift or undergo dynamic changes during intervention, complicating the precise localization
of robots, especially at small scales. Integrating robotic systems with clinical imaging modalities, such as X-
ray fluoroscopy [78,94] , ultrasound [57,77,212] , and magnetic resonance imaging [213,214] , is essential for establishing
robust, real-time, closed-loop control. Furthermore, advances in image registration, multimodal sensor