The work was carried out by Li Ma, Bingyue Yang, Jie Shao, Qiao Hu and Zicai Zhu of the School of Mechanical Engineering at Xi’an Jiaotong University, working with the Shaanxi Key Laboratory of Intelligent Robots and the Department of Surgical Oncology at the university’s First Affiliated Hospital. Their motivation stems from a long-standing tension in minimally invasive surgery. Since the laparoscopic revolution of the 1990s, surgeons have operated through narrow keyhole incisions using long rigid instruments that dramatically reduce patient trauma compared with open surgery. But that reduction in invasiveness carries a steep mechanical penalty. Rigid laparoscopic tools pass through the abdominal wall at a fixed pivot point, which reverses the direction of instrument motion at the tip, restricts the wrist-like freedom a surgeon naturally expects, and forces uncomfortable, fatigue-inducing postures during long operations.

Full-scale surgical robots such as the da Vinci system solved much of that problem by decoupling the surgeon’s hands from the instrument, filtering tremor and restoring wrist-like articulation deep inside the body. Decades of clinical data and large single-institution series, including reports of more than 10,000 robotic cases, have demonstrated the safety and effectiveness of such platforms, particularly for complex procedures like gastric and colorectal cancer resection. Yet these systems remain extraordinarily expensive to purchase and maintain, occupy significant operating room space, and require dedicated teams of scrubbed and unscrubbed staff to set up and dock. For many hospitals, particularly in resource-constrained health systems, the cost calculus simply does not close. Handheld robotic instruments emerged as a counter-proposal: keep the surgeon at the table with direct visual and manual contact, but embed the articulation and dexterity of a robot into the instrument itself.

Earlier handheld devices pursued that goal largely through clever mechanics rather than electronics. Instruments such as FlexDex translated wrist and finger motion mechanically, through cable linkages and compliant mechanisms, into tip articulation without any motors or computers at all, offering what researchers have described as robot-like dexterity without computers and motors. Other commercial and academic designs, including multi-degree-of-freedom articulating instruments, have shown that challenging suturing tasks become easier when the tip can flex. But mechanical solutions impose compromises. The mapping between the surgeon’s hand and the instrument tip is fixed at the time of design, wrist-controlled operation can feel unintuitive during fine maneuvers, and the mechanics can be complex and difficult to manufacture. The Xi’an team chose a different route: put small electric motors in the handle and let software define the relationship between hand motion and tip motion.

The resulting prototype is organized around three key subsystems. The first is the joystick controller held under the surgeon’s fingers. Its geometry is deliberately isomorphic with the continuum end effector at the instrument tip, meaning the joystick’s axes of motion correspond directly, in both direction and sense, to the rotations of the tip. When the surgeon tilts the joystick forward, the tip pitches forward inside the body; when the joystick rolls left, the tip yaws left. That one-to-one correspondence is the essence of master-slave isomorphism, and it stands in contrast to the mirrored, reversed kinematics that plague conventional rigid laparoscopic instruments. Intuitive control matters not merely for comfort but for safety, because cognitive load during surgery is a finite resource, and every moment spent mentally inverting a coordinate system is a moment not spent on the anatomy.

The second subsystem is the end effector itself, a multi-layer cross-shaped Hooke hinge structure that behaves as a continuum wrist. Rather than using a small number of discrete rotational joints, the device bends continuously through stacked hinge layers arranged in orthogonal orientations, allowing the tip to combine yaw and pitch into smooth, compound trajectories. The researchers modeled this behavior using continuous curvature methods, a mathematical framework standard in continuum robotics that describes the backbone of the manipulator as a curve with smoothly varying curvature rather than as a chain of rigid links. This modeling choice is what enables the joint incremental proportional control scheme they propose: instead of commanding absolute joint positions, the controller computes small increments for each of the tip’s drive joints proportional to the instantaneous error between the desired and current tip orientation, producing stable, smooth tracking even across the full ±90-degree combined range.

The third subsystem is a quick-change mechanism that addresses the practical economics of surgery. Surgical instruments are consumables; tips dull, foul, and must be replaced between uses or between steps of a procedure. In conventional robotic setups, instrument exchange can consume valuable operating room minutes. The team’s mechanism lets the sterile tip module be detached and replaced in less than seven seconds, a figure they measured directly on the prototype. Because the electronics and motors reside in the reusable handheld master unit, only the inexpensive tip module is discarded, an architecture intended to cut per-procedure costs substantially compared with robotic platforms whose instrument arms are themselves expensive limited-use items.

Quantifying how faithfully the slave tip follows the master joystick required an independent measurement system, so the team constructed an optical three-dimensional motion capture rig to record both the joystick trajectory and the corresponding tip trajectory simultaneously. From those synchronized recordings they extracted the device’s accuracy and latency characteristics. The maximum trajectory error across the tested movements was 3.43 millimeters, and the master-slave control delay was 0.304 seconds. Latency is a critical parameter for any teleoperated or semi-teleoperated surgical device; prior research using surgical simulators has examined how delay degrades performance in telesurgery and sought to establish acceptable latency thresholds, since even modest lag disrupts the sensory-motor loop between what a surgeon sees and what the instrument does. A delay of roughly a third of a second places the prototype in a range where deliberate surgical maneuvers remain feasible, though the authors note it as a characteristic of the current early-stage prototype rather than an optimized value.

The ±90-degree combined yaw and pitch range deserves particular attention because it roughly doubles the articulation available through many existing handheld wrist instruments and approaches the freedom that full robotic platforms provide. The team demonstrated that the robot can execute multi-axis circular movements with excellent consistency, meaning that when commanded to trace circular paths that combine yaw and pitch simultaneously, the tip reproduced those compound trajectories repeatably. Consistency of this kind matters clinically because suturing in confined spaces, for example deep in the pelvis or around the esophageal hiatus, frequently demands oblique, compound wrist angles rather than simple planar motion. Finally, in proof-of-concept trials with a laparoscopic simulator, the device grasped tissue and performed suturing tasks more easily than would be expected with conventional rigid instruments, providing an early functional validation that the kinematic numbers translate into practical surgical capability.

Beyond its immediate performance figures, the study makes a methodological contribution. The joint incremental proportional control approach, in which each rope-driven joint of the continuum mechanism is controlled through proportional increments derived from the master input, is general enough, the authors argue, to be applied to other rope-driven continuum robots. Tendon-driven continuum architectures are now widespread in medical robotics, from single-port access systems to steerable catheters and endoscopes, and they share a common control challenge: the mapping between actuator cable lengths and tip pose is nonlinear, coupled, and sensitive to friction and compliance in the transmission. An incremental proportional scheme sidesteps some of the calibration burdens of model-based approaches by reacting to error locally at each joint, making it an attractive, computationally lightweight option for compact, battery- or motor-driven devices where embedded processing power is limited.

The device’s modular, handheld philosophy also speaks to a broader conversation about how surgical robotics should scale. Several research groups have previously explored dexterous and modular handheld surgical robots, and commercial handheld articulating instruments have accumulated growing clinical datasets, including multicenter analyses of learning curves for procedures such as laparoscopic right hemicolectomy. The consistent finding across that literature is that wrist-like articulation shortens the time it takes surgeons to master difficult laparoscopic maneuvers and reduces the ergonomic strain that contributes to chronic injury among minimally invasive surgeons. A handheld electric device that matches those ergonomic benefits while avoiding the capital cost of a full robotic platform could widen access to advanced minimally invasive care in hospitals and health systems that cannot justify a docked robot for their surgical volumes.

The Xi’an work remains a prototype study, and the authors are careful about its scope. All testing to date has been benchtop and simulator-based; the study involved no patients or live animals and required no ethics committee approval. Achieved control delay, trajectory error, and the durability of the quick-change mechanism under real sterilization cycles and real tissue mechanics remain to be validated in more demanding preclinical settings. Latency in particular will need to fall if the device is to support the fastest phases of surgical work. Nevertheless, the combination of a full ±90-degree articulation envelope, sub-3.5-millimeter trajectory error, sub-seven-second instrument exchange, and an isomorphic finger-controlled interface represents one of the most complete engineering demonstrations to date of what a genuinely affordable, electric handheld surgical robot can achieve. The research was supported by the Postdoctoral Fellowship Program of the China Postdoctoral Science Foundation and the Postdoctoral Research Funding Project of Shaanxi Province. If subsequent iterations close the latency gap and survive preclinical testing, the operating room of the future may look less like a stage dominated by a four-armed tower and more like a surgeon at the table, holding in one hand a robot that fits where a scalpel does.

Historically, robot-assisted surgery has relied on specialized platforms purpose-built for precise tasks, such as Intuitive Surgical’s da Vinci system. However, the degree to which modern humanoid robots can match or approach the rigorous control, accuracy, and safety standards for minimally invasive surgery has remained uncertain. Addressing this gap, a recent study systematically evaluates contemporary humanoid robots’ aptitude for laparoscopic surgery through a comprehensive framework encompassing teleoperated control using general-purpose instruments.

This pioneering research incorporated multiple experimental tiers: benchtop system characterization, dry-laboratory trials involving users with varied surgical expertise, and in vivo experiments using porcine models. These methodical assessments quantified the humanoids’ technical capabilities, operational performance in surgical tasks, and readiness for clinical environments compared to established robotic surgical platforms.

Findings reveal that humanoid robots exhibit promising dexterity and control potentials, able to manipulate laparoscopic tools with precision under teleoperation. However, several technical hurdles remain, including optimizing actuator responsiveness for rapid adjustments, refining whole-body control algorithms to handle complex articulated motions, and ensuring consistent haptic feedback to improve operator perception during interventions. The research underscores the criticality of integrating advanced model-predictive control and innovative learning-based approaches to bridge these gaps.

Moreover, the experiments highlight the robots’ handling of dynamic intraoperative conditions, demonstrating their capacity to perform essential surgical maneuvers while maintaining stability in unpredictable settings. The porcine in vivo results particularly validate the feasibility of humanoid-based systems to assist in minimally invasive procedures without compromising safety or accuracy.

Despite the encouraging outcomes, the study emphasizes that current humanoid robots are not yet fully mature for direct clinical deployment. Further technical refinements and extensive validations remain essential. The ability to safely and effectively collaborate with human surgeons, adapt to diverse intraoperative scenarios, and achieve the finesse required for intricate tissue manipulation are paramount goals that future research must target.

This investigation represents a milestone in assessing how general-purpose humanoid robots can be adapted for high-stakes medical roles. It delivers an evidence-based roadmap outlining both the capabilities realized and the challenges ahead, shaping the trajectory of robotic technologies poised to transform surgical care.

As healthcare systems worldwide grapple with workforce shortages and increasing surgical demands, the integration of humanoid robots offers a transformative vision—augmented surgical teams equipped with versatile robotic collaborators, enhancing precision, safety, and accessibility of minimally invasive surgery.

Subject of Research: Humanoid robot capabilities in laparoscopic surgery
Article Title: In vivo feasibility study of humanoid robots in surgery
Article References: Liang, Z., Thareja, N., Zhang, P. et al. In vivo feasibility study of humanoid robots in surgery. Nature (2026). https://doi.org/10.1038/s41586-026-10796-x
DOI: https://doi.org/10.1038/s41586-026-10796-x

Conventionally, robotic arms are perceived as machines with rigid segments and joints that execute rotational or linear motions. These designs, iconic in cultural representations such as the Transformers series, fall short when tasked with navigating highly confined spaces replete with soft tissues or fragile components. In scenarios like internal surgical procedures or inspections within compact machinery, the rigidity of traditional robots limits effectiveness and increases the risk of collateral damage.

To overcome these limitations, continuum robots (CRs) have been developed, characterized by their flexible, snake-like bodies that conform dynamically to their surroundings. Among CRs, the tendon-driven continuum robot stands out for its simplicity, lightweight architecture, and high degrees of maneuverability. Its design comprises a flexible backbone actuated by multiple thin tendons that run longitudinally alongside the robot’s structure. By selectively tensioning these tendons, the robot can bend, twist, and articulate smoothly in numerous directions, closely mimicking the natural movements found in biological appendages such as octopus tentacles or elephant trunks.

Despite the intuitive mechanical design, the control of TDCRs presents a significant analytical challenge. Unlike traditional robots with a finite, often limited number of joints, TDCRs exhibit theoretically infinite degrees of freedom, making precise navigation a complex, computationally intensive problem. The complexity escalates further with robots composed of multiple sections, where the actuation of one tendon might inadvertently influence the motion of adjacent sections, resulting in a tightly coupled control problem.

Addressing this formidable challenge, the IITGN research team introduced the concept of the virtual actuation space (VAS), a transformative approach to streamline the control of multi-section TDCRs. Rather than directly commanding the physical tendons, VAS abstracts the robot’s movement into a conceptual two-parameter representation: the direction of bending and the magnitude of curvature for each segment. This simplification decouples the control mechanism, allowing each section of the robot to operate independently without undesired interference from others—a significant advancement over traditional tendon-interconnected control strategies.

The virtual actuation space reduces the computational overhead typically required for real-time motion planning, enabling more responsive and precise robot movement. This methodology compensates for the complex interdependencies within multi-section TDCRs by adjusting actuation parameters in a virtual coordinate system, which subsequently maps back to tendon displacements. Through this abstraction, the control problem transforms from an unwieldy infinite-dimensional system into a manageable finite parameter space.

Validating the efficacy of the VAS approach involved the creation of a prototype robotic arm comprising two independently actuated sections. Each section was controlled by six motors capable of finely adjusting tendon lengths, facilitating nuanced bending profiles. To quantify performance, the researchers employed high-resolution motion capture technology, outfitting the robot with small LED markers whose positions were tracked in three-dimensional space. This setup allowed for precise feedback, with the control algorithm continually correcting motor movements to align the robot’s actual position with its designated trajectory.

Extensive experiments demonstrated the robot’s ability to execute complex trajectories with remarkable precision. In one test, the robot’s tip traced the vertices of a pentagon sequentially, achieving an error margin below one percent—an extraordinary feat in continuum robot control. Additional trajectories mimicked the shapes of a two-petalled flower, spiral, circle, and arbitrary curves, confirming the system’s versatility. Particularly noteworthy was the independent operation of the robot’s sections—one section could bend while the other remained straight, highlighting the decoupled control capability enabled by VAS.

The implications of this research extend deeply into practical applications requiring dexterous, reliable robotic manipulators. In surgical contexts, the ability of a TDCR to maneuver delicately without cross-section interference could elevate the safety and efficacy of minimally invasive procedures. Beyond healthcare, industries such as aerospace and manufacturing stand to benefit from robotic systems capable of inspecting and maintaining equipment nestled within confined and complex assemblies. The adaptability and reduced computational demand promise scalable solutions for multi-sectional continuum robots, accommodating more intricate designs as technological demands evolve.

This breakthrough aligns with India’s national directives aimed at propelling the country to the forefront of robotics innovation by 2030. By integrating cutting-edge research with strategic governmental initiatives like the National Strategy on Robotics and Make in India 2.0, IITGN’s work embodies a fusion of academic excellence and societal impact. The institute fosters a dynamic culture around robotics, inspiring students and researchers to explore and innovate in this transformative domain.

The researchers have secured intellectual property protection through a patent (application number 202421002550) filed with the Indian Patent Office, underscoring the novelty and commercial potential of the VAS-based control method. They acknowledge crucial financial support from the Gujarat Council on Science and Technology and appreciate collaborative insights from faculty colleagues and team members within their robotics lab.

In essence, the virtual actuation space framework represents a paradigm shift in continuum robot control, bridging the divide between mechanical complexity and computational tractability. Its ability to enhance precision, reduce control interdependencies, and maintain real-time responsiveness marks a milestone in the evolution of flexible robotics. As this technology matures and proliferates across domains, it promises to unlock new frontiers where delicate, adaptable, and precise robotic movement is indispensable.

Subject of Research: Control methodologies for tendon-driven continuum robots
Article Title: Trajectory tracking of multi-section tendon-driven continuum robots using virtual actuation space control
News Publication Date: 27-Feb-2026
Web References: DOI link
Image Credits: Credit to authors and the IITGN Robotics Lab team at the Indian Institute of Technology Gandhinagar, Gujarat, India

Keywords

Tendon-driven continuum robot, tendon actuation, robotic manipulation, flexible robots, virtual actuation space, robotics control, multi-section robot, computational efficiency, surgical robotics, precision robotics, motion capture, robotic trajectory tracking

At the forefront of this research is the acute realization of the potential that micro and nanorobots have in revolutionizing biomedicine. The applicability of these robots extends to numerous fields, particularly in areas such as minimally invasive surgeries, targeted drug delivery, and therapeutic interventions. Previous endeavors in the realm of microrobotics have demonstrated successful targeted delivery mechanisms using drug-loaded biohybrid microrobots, underscoring their transformative potential. However, the functionality of these microswarms has often faced substantial hurdles, primarily related to achieving autonomous navigation that is both efficient and collision-free in the dynamically changing landscapes of confined spaces.

One critical challenge that researchers encounter in microswarm navigation is the inherent variability in size among microrobots. Traditional path-planning techniques frequently neglect to account for the physical dimensions of these robots, leading to inadequate solutions. Additionally, the reconfigurability of deformable microrobots poses another layer of complexity that existing algorithms fail to adequately address. Recognizing these issues, the research team devoted their efforts toward developing a cutting-edge deep learning-based scheme specifically designed to enhance navigation capabilities in microswarm deployments.

The nucleus of this innovative framework is a deep Q-network (DQN) that specializes in path planning. Utilizing reinforcement learning principles, this approach excels in real-time planning, making it particularly adept at handling complex scenarios characterized by channel-like structures and obstacles that may shift dynamically. Rigorous training on various sizes of microrobots, including configurations measuring 1×1, 2×2, 3×3, and 5×5 pixels, resulted in an impressive success rate exceeding 92% after completion of 30,000 training episodes. This efficacy not only proves the validity of the model but also showcases its capability to recommend adaptive safe distances to accommodate varying robot sizes, thus ensuring reliable navigation even in cluttered environments.

The research team further expanded upon their DQN-based approach by introducing a pattern-distribution planner focused on real-time swarm pattern planning and control. This planner employs an intelligently constructed cost function to pinpoint the optimal distribution of the swarm, facilitating seamless maneuvering amidst surrounding obstacles. Leveraging a deep convolutional neural network (DCNN), the researchers demonstrated the capability to generate optimal swarm patterns. In scenarios necessitating agility, such as navigating ribbon-like formations, the DCNN adeptly calculated suitable swarm distributions by factorizing in the presence of changes in the environment.

This comprehensive DL-based framework not only surpasses conventional planning methodologies in dynamic and variable situations but also exemplifies its remarkable adaptability to different microrobot sizes. What this indicates is that such navigation strategies could provide reliable, efficient solutions for a diverse array of microswarm applications, thus transforming the way we envision their deployment across medical, industrial, and agricultural fields.

The implications of this research extend far beyond mere academic exercise. Anticipating the needs of practical application, the study paves the way for a deeper understanding of how these microswarms might autonomously operate in more complex ecosystems, such as physiological surroundings or environments complicated by hydrodynamic disturbances. The findings hold the potential to crystallize new avenues for microswarm applications in biomedicine, offering enhanced therapeutic approaches and innovative surgical techniques that can mitigate risks associated with traditional methods.

To encapsulate the crux of the findings, the research titled “A Deep Learning-based Framework for Environment-adaptive Navigation of Size-adaptable Microswarms” serves as an important reference point for future exploration within the field. The work represents a convergence of artificial intelligence and robotics, showcasing how computational advancements can bridge the chasm between theoretical modeling and practical application. It stands to incite further inquiry into the autonomy of microswarms, with prospects that extend into interactive and responsive biomedicine.

In summary, the research published in Engineering underscores the incredible potential of deep learning technologies as tools for autonomous navigation in microswarms. As teams worldwide continue to explore the remarkable capabilities of microrobots, studies such as this will undoubtedly be vital. The proactive acquisition of new knowledge surrounding the robotics field will only serve to enhance our capabilities, ultimately paving the way for significant advancements that could redefine our approach to complex biological challenges.

With this framework, researchers and practitioners in the field are encouraged to think creatively about the possibilities that lie ahead. The integration of deep learning solutions into the practical domain of microswarm navigation heralds a new era of exploration and application that could enrich the field of microscale robotics and expand its implications across various industries.

The significance of this research cannot be understated. As we stand on the brink of technological advances in microrobotics and artificial intelligence, the work of this research team showcases the transformative possibilities available to us. The developments articulated within this paper will set the tone for subsequent investigations, ensuring that microswarms are not only a product of scientific imagination but a practical reality poised to make waves in numerous fields.

In conclusion, the collaboration and innovation showcased in this study illuminate the path forward for integrating deep learning into autonomous robotics. The potential benefits not only address existing limitations but also create new opportunities for research and development, paving the way for future enhancements in microswarm technologies.

Subject of Research: Environment-adaptive navigation of size-adaptable microswarms
Article Title: A Deep Learning-based Framework for Environment-adaptive Navigation of Size-adaptable Microswarms
News Publication Date: 2-Dec-2024
Web References: DOI Article
References:
Image Credits: Credit: Jialin Jiang et al.

Keywords

Navigation, Deep Learning, Microrobots, Microswarms, Path Planning, Biomedicine, Reinforcement Learning.