The core insight of the new work is that no single controller suits every operating regime of a self-balancing wheelchair. Near the upright equilibrium, the system behaves almost like a linear inverted pendulum, and a linear quadratic regulator, or LQR, is an ideal tool: it is computationally cheap, mathematically well understood, and delivers smooth, accurate regulation when deviations are small. But when the chair is tilted far from vertical, say 35 to 45 degrees, the linear approximation breaks down badly. The nonlinear terms in the equations of motion, particularly the sine of the tilt angle and the coupling between translation and rotation, dominate the behavior, and an LQR designed around the upright point loses authority rapidly. Nonlinear model predictive control, or NMPC, can handle these large deviations because it re-solves an optimization problem at every sampling instant using the full nonlinear dynamics, but that repeated online optimization carries a heavy computational price and is not always superior to simpler methods once the system is already close to equilibrium.

The hybrid NMPC–LQR scheme resolves this tension with an elegant switching law. Whenever the magnitude of the pitch angle exceeds a predefined switching threshold, the NMPC controller takes command, exploiting its nonlinear recovery capability to drive the chair back toward upright under explicit constraints on the available wheel torque. Once the tilt angle falls below the threshold, control authority is handed to the LQR, which stabilizes the system efficiently in the near-equilibrium region. The result, according to the team’s simulations, is a controller that recovers faster than either standalone method while cutting the average computation time from 48.04 minutes for standalone NMPC to 35.00 minutes for the hybrid scheme across repeated simulation runs, a meaningful reduction in the online optimization burden that matters for real-time embedded implementation.

Choosing the switching threshold is not a trivial matter, and the researchers devoted an entire methodological framework to it. If the threshold is set too small, NMPC remains active until the chair is nearly upright, wasting computational resources and delaying the efficient local regulation that LQR provides. If it is set too large, the controller hands over to LQR while the system is still in a strongly nonlinear region, where the linear controller’s assumptions are invalid and recovery performance degrades. The team first performed a linearization-error analysis, comparing the nonlinear model against its linearized counterpart across angular ranges of 0.015 to 0.150 radians, and found that the mismatch in pitch angular acceleration grows gradually within that interval but rises sharply beyond it, establishing 0.015 to 0.150 radians as a defensible search range.

Within that range, the researchers evaluated six candidate thresholds using a normalized five-index criterion that combines settling time, steady-state error, steady-state standard deviation, drift rate, and maximum steady-state deviation into a single composite cost. The weights were deliberately chosen to prioritize steady-state accuracy, since precise balance near upright is the primary requirement for occupant safety and comfort. Among the candidates tested under nominal model parameters, a threshold of 0.100 radians yielded the lowest composite cost, representing the best trade-off between transient recovery and steady-state regulation. The team went further with a sensitivity analysis, varying the equivalent upper-body mass and pendulum length independently, and discovered a clear empirical relationship: heavier occupants require smaller switching thresholds, so that NMPC stays in charge over a wider angular range. The fitted relation, with a coefficient of determination of 0.945, offers designers a practical rule for adapting the threshold to different users.

The modeling foundation beneath all of this is a reduced-order description of the wheelchair–occupant system derived through the Lagrange method. The chair and its occupant are treated as a rigidly connected inverted pendulum, with the two driving wheels assumed identical and in pure rolling contact with the ground. Nonholonomic rolling constraints are eliminated using a null-space projection, which removes the constraint reaction forces from the equations and yields a compact two-degree-of-freedom model in longitudinal displacement and pitch angle. The authors are candid about the limits of these assumptions: yaw, roll, tire deformation, wheel slip, and active occupant motion are not modeled, and the rigid-body assumption holds only when the occupant sits relatively still. Rather than pretending these effects do not exist, the team treated them as unmodeled disturbances and stress-tested the controller against them.

Those stress tests form the empirical heart of the study. In comparative simulations with initial tilt angles of 35, 40, and 45 degrees, the hybrid controller achieved the shortest settling time under all three conditions, and at 45 degrees it also delivered the smallest steady-state error, standard deviation, and maximum deviation. Against a sliding-mode control benchmark, a popular robust nonlinear method, the contrast was stark: the sliding-mode controller failed to recover from 40 and 45 degree tilts, while the hybrid controller remained stable across every tested initial angle. Disturbance-rejection trials told a similar story. Under pulse disturbances of 50 to 80 newtons, the chair exhibited a secondary deflection but never diverged, and under persistent step disturbances of 10, 20, and 30 newtons the peak deviations grew to 3.62, 5.89, and 8.11 degrees respectively, yet always remained bounded and returned toward equilibrium.

Robustness to real-world imperfections was examined through flat-ground friction tests in which the static friction coefficient was varied from 0.03 to 0.30, spanning low-, medium-, and high-friction contact conditions. The wheelchair did not overturn under any of them, although the high-friction case produced larger residual pitch oscillations, suggesting that stronger wheel–ground interaction can amplify small fluctuations near equilibrium. The authors interpret these results as bounded recovery rather than exact convergence, an honest framing that reflects the difficulty of the problem. Finally, the team validated the strategy on a desktop-scale physical prototype built on the WHEELTEC B570 architecture, using an STM32F103C8T6 microcontroller and an MPU6050 inertial sensor. In balance-recovery experiments starting from 25, 30, and 35 degree tilts, the prototype returned to a small neighborhood of upright in every case, with transient oscillations that grew as the initial angle increased but never threatened stability.

The significance of this work extends beyond wheelchairs. Two-wheeled inverted-pendulum systems appear in segways, delivery robots, and personal mobility devices, and the challenge of balancing fast nonlinear recovery against efficient local stabilization is universal. What distinguishes this study is its methodological rigor in quantifying the handover point between controllers: rather than picking a switching threshold by intuition, the team grounded it in linearization-error analysis, optimized it with a multi-index criterion, and characterized how it should shift with user mass. That kind of disciplined threshold selection addresses a recognized gap in the hybrid control literature, where switching boundaries are often chosen heuristically and their parameter dependence is rarely measured.

Limitations remain, and the authors enumerate them plainly. The model is planar and does not capture yaw, roll, lateral instability, or the coupled dynamics of a moving human body. The stability argument rests on local LQR analysis and extensive closed-loop validation rather than a rigorous global Lyapunov proof for the switched nonlinear system, which remains an open theoretical problem. The prototype is desktop-scale, and full-scale trials on slopes and uneven terrain are still to come. Future work, the team says, will pursue human–wheelchair coupled modeling, adaptive threshold calculation, explicit friction compensation, and robust NMPC with terminal constraints. For now, the study offers a concrete, tested blueprint for making self-balancing wheelchairs safer under exactly the conditions that frighten their users most: a sudden jolt, a heavy passenger, or a chair tipped dangerously far from vertical.

Subject of Research: Hybrid nonlinear model predictive and linear quadratic balance control for two-wheeled self-balancing wheelchairs

Article Title: Hybrid nonlinear model predictive and linear quadratic balance control for a two-wheeled self-balancing wheelchair

Article References: Sun, H., Zhang, X., Gu, Y., Wang, S., Chen, W., Zhang, S., Xu, T., Yu, H., & Meng, Q. (2026). Hybrid nonlinear model predictive and linear quadratic balance control for a two-wheeled self-balancing wheelchair. Mechanical Sciences, 17(2), 713-729. https://doi.org/10.5194/ms-17-713-2026

Image Credits: AI Generated

DOI: 10.5194/ms-17-713-2026

Keywords: self-balancing wheelchair, model predictive control, LQR, nonlinear control, inverted pendulum, switching threshold, disturbance rejection, assistive robotics, robustness, prototype experiments, rehabilitation engineering, underactuated systems

The motivation is practical and pressing. Nursing staff spend significant time and physical effort making beds, and the repetitive bending, lifting, and reaching involved contributes to musculoskeletal strain. Automated bed-making could reduce that physical burden while also lowering infection risk, since consistent, standardized bed preparation matters in clinical environments. Surveys cited in the study suggest that nurses and patients are generally willing to work alongside service robots in healthcare settings, and recent studies indicate that nursing robots can reduce workload in general adult wards. But translating that demand into a working machine requires solving a problem that has resisted a decade of research: reliably manipulating cloth, the archetypal deformable object, in an unstructured real-world environment.

The core of the new framework is a formal twelve-step task decomposition of bed-making. Rather than treating the job as one monolithic skill, the researchers broke it into a reproducible sequence of task-level stages, bridging high-level semantic reasoning and low-level manipulation control. This structured representation matters because long-horizon tasks compound errors: a small mistake in an early stage, such as a poor grasp of a sheet corner, propagates through every subsequent action. By formalizing the task as a discrete sequence, the team created a framework in which each step can be monitored, evaluated, and recovered from independently, and in which performance can be measured stage by stage rather than only as an overall pass-or-fail outcome.

Perception is driven by semantic keypoints, a technique that has become a powerful tool for handling deformable objects. Instead of asking a vision system to label every pixel of a crumpled sheet, the robot uses vision-language models to localize semantically meaningful points: the corner of a sheet, the edge of a mattress, the midpoint of a fold. The approach builds on open-vocabulary detection systems such as Grounding DINO and YOLO-World, and on the visual grounding tradition established by CLIP, allowing the robot to identify task-relevant points from natural language descriptions without needing task-specific retraining. Earlier work on general-purpose clothes manipulation with semantic keypoints demonstrated the promise of this representation, and the new study extends it to the more demanding setting of dual-arm bed-making, where two manipulators must act on the same deformable object in a coordinated way.

Once keypoints are localized, the system plans coordinated dual-arm motion for fabric handling. Two arms are not merely a convenience here; they are essential. Spreading a sheet across a mattress, aligning its edges, and folding it flat all require simultaneous control of two distant points on the same cloth, with continuous tension maintained between them. The manipulation planner generates motions that respect the coupled geometry of the fabric, drawing on a research lineage that includes model-free visual servoing for deformation control, dynamic movement primitives for deformable manipulation, and learned pick-point detection for robotic bed-making. The team also employed hand-eye calibration using 3D-to-3D correspondences to align what the cameras see with where the arms actually move, a deceptively mundane step that is critical for millimeter-level grasp accuracy.

Perhaps the most consequential engineering contribution is the closed-loop recovery mechanism. In open-loop systems, a robot executes a planned sequence and hopes for the best; any failure, such as grasping two layers of fabric instead of one, silently corrupts everything that follows. The new system instead monitors each step, detects unsuccessful grasps, and updates its perception, planning, and execution accordingly. If a grasp fails, the robot re-perceives the scene, re-plans, and retries, treating uncertainty as an expected condition rather than an exception. This probabilistic stance toward real-world execution, long a theme in robotics, is what lifts the system’s average step-wise success rate to 91.9 percent across the twelve stages, a figure that would be unattainable with naive open-loop execution.

The experimental evaluation on a real robotic system reveals a striking asymmetry in difficulty. Rigid manipulation tasks, such as moving objects with well-defined geometry, achieved near-perfect reliability. Deformable operations, particularly sheet folding, remained the primary bottleneck, hampered by perception and grasp uncertainty. The most troublesome failure mode identified by the researchers occurs in the early stages, when the robot must distinguish individual fabric layers. Grabbing two layers of a sheet when only one is intended may look superficially successful, but the error cascades through the task, distorting subsequent folds and alignments in ways that later recovery steps cannot fully undo. This error propagation across task stages is precisely why the overall task success rate lags behind the per-step figure, and it encapsulates the central challenge of long-horizon deformable manipulation.

The study situates itself within a rapidly evolving research landscape. Deep-learned pick-point detection for robot bed-making dates back to work presented at the International Symposium on Robotics Research in 2019, and simulation-to-real reinforcement learning for deformable objects has been explored since at least 2018. More recently, large language models have been used for symbolic planning in long-horizon deformable assembly, and open-source vision-language-action models such as OpenVLA have begun to unify perception and control. Tactile sensing is emerging as a complementary channel: researchers have shown that robots can singulate layers of cloth using tactile feedback, and transformer-based multimodal frameworks are being developed for material perception. The Taiwanese team’s contribution is to integrate these threads, vision-language keypoint localization, dual-arm planning, and closed-loop recovery, into a single working system tested end-to-end on hardware, rather than in simulation alone.

The authors are candid about the limitations and the path forward. Future work will integrate tactile and visual feedback to improve layer-state estimation, giving the robot a direct physical sense of how many fabric layers it holds, and to enable adaptive recovery strategies that respond to the actual state of the cloth rather than only its appearance. This is a sensible direction, since vision alone struggles with the self-occlusion and ambiguity that make layered fabric so treacherous. The research was supported by Taiwan’s National Science and Technology Council and the Industrial Technology Research Institute, reflecting an institutional commitment to translating assistive robotics from laboratory demonstrations into clinically useful systems.

For the field of embodied AI, the 91.9 percent step-wise success rate is both an achievement and a diagnostic. It shows that structured task decomposition, semantic perception, and failure-aware control can carry a robot through a genuinely long-horizon household task on real hardware. It also shows, with unusual clarity, exactly where the frontier lies: not in rigid grasping, which is essentially solved at this scale, but in the entangled physics of cloth, where seeing is not always knowing. If the next generation of these systems can feel the difference between one layer and two, the humble act of making a bed may become one of the first full-scale deformable manipulation tasks that robots perform reliably alongside nurses, and a template for the laundry-folding, dressing-assisting, and linen-handling robots that would follow.

Subject of Research: Autonomous dual-arm robotic bed-making using semantic keypoint perception and closed-loop recovery for deformable object manipulation

Article Title: Dual-arm manipulation driven by semantic keypoints for autonomous bed-making

Article References: Shih, C.-H., Cheng, P.-H., & Chuang, Y.-Y. (2026). Dual-arm manipulation driven by semantic keypoints for autonomous bed-making. International Journal of Intelligent Robotics and Applications. https://doi.org/10.1007/s41315-026-00597-w

Image Credits: AI Generated

DOI: 10.1007/s41315-026-00597-w

Keywords: assistive robotics, deformable object manipulation, semantic keypoints, dual-arm manipulation, vision-language models, closed-loop recovery, long-horizon planning, hospital automation, robotic perception, fabric handling, embodied AI, nursing robots

Millimetre-wave radar, the same class of technology used in automotive collision avoidance and next-generation wireless communications, emits radio waves in the millimetre band and listens for reflections. When those reflections bounce off a moving person, the radar’s signal processing pipeline produces a point cloud: a scattered set of three-dimensional coordinates tracing the surfaces and limbs that returned the signal. Unlike the dense, well-structured point clouds produced by lidar or depth cameras, radar point clouds are notoriously sparse, often containing only a handful of points per frame, and they are contaminated by multipath reflections and hardware noise. That sparsity and irregularity is precisely what has made them so difficult for deep learning systems to interpret.

The Aston team’s insight begins with a widely used building block of point-cloud deep learning called EdgeConv, introduced originally in the Dynamic Graph CNN architecture. EdgeConv treats each point as a node in a graph, connects it to its k nearest neighbours, and computes features along the edges connecting them, capturing the local geometry of the scene. The problem, as the researchers point out, is that conventional EdgeConv is deterministic and indiscriminate: it always uses every one of its k nearest neighbours in every forward pass, faithfully incorporating noisy, redundant or uninformative points into the representation. On clean, dense benchmark datasets from 3D vision, that determinism is a virtue. On sparse, noisy radar data, it becomes a liability, because the same corrupted neighbourhoods poison the features pass after pass.

SEdgeNet’s central contribution is a redesigned operator the authors call Stochastic Edge Convolution, or SEdgeConv, which is built from two complementary mechanisms. The first, Stochastic Neighbour Sampling, randomly selects only a subset of each point’s k nearest neighbours during each forward pass, governed by a sampling ratio that determines what fraction of the neighbourhood is used. The idea echoes dropout-style regularisation techniques from the graph neural network literature, such as DropEdge: by forcing the network to produce useful representations from constantly varying, randomly drawn subsets of neighbours, the model is discouraged from leaning on any single noisy point or fixed local configuration. Crucially, the researchers note, this is not a permanent loss of information. Across the many forward passes of training, the full neighbourhood structure remains accessible; the stochasticity simply prevents overreliance on any particular slice of it.

The second mechanism, Edge-wise Factorised Convolution, tackles the computational side of the problem. Standard EdgeConv applies a dense transformation across all edge feature channels jointly, which is expensive. EFC instead adapts the depthwise separable convolution paradigm, originally developed to make convolutional networks efficient on regular image grids, to the irregular edge features produced by stochastically sampled neighbourhoods. Each channel of the edge feature tensor is first transformed independently, and the per-channel responses are then fused through a pointwise linear projection. The authors stress that the efficiency gain comes from coupling this factorised processing with the dynamically varying graph structure created by the stochastic sampling, rather than from the factorisation scheme alone. Neighbour features are then aggregated with symmetric max pooling, yielding a permutation-invariant representation that does not depend on the ordering of points.

Around these operators, SEdgeNet stacks multiple SEdgeConv layers and concatenates their outputs along the channel dimension, forming a multi-scale embedding that preserves both fine-grained local geometry and more abstract global structure. A dual pooling strategy, combining adaptive max pooling with adaptive average pooling, distils this embedding into a fixed-length global descriptor; max pooling emphasises salient structures while average pooling encodes overall shape statistics. The pooled vectors feed a three-layer multilayer perceptron classifier that outputs the predicted activity class. The model is trained with stochastic gradient descent, a cosine annealing learning rate schedule, cross-entropy loss and early stopping, with all reported results averaged over three independent training runs using different random seeds.

The team evaluated SEdgeNet on two public mmWave radar datasets. The first, MMActivity, contains 93 minutes of point cloud data covering five activities performed by two subjects, captured with a Texas Instruments IWR1443BOOST sensor and processed into 12,097 training and 3,538 test samples using two-second windows. The second, MiliPoint, is far more demanding: 49 activities performed by 11 subjects, recorded with a TI IWR1843 radar and split 80-10-10 for training, validation and testing. On MMActivity, SEdgeNet outperformed the best previous model by 0.72 percent in accuracy, 0.71 percent in precision, 2.7 percent in recall and 0.71 percent in F1 score. On MiliPoint, the gains were larger still: 5.42 percent in accuracy, 4.91 percent in precision, 4.81 percent in recall and 5.26 percent in F1 score, establishing new state-of-the-art results on both benchmarks.

What makes those numbers remarkable is that they were achieved while using fewer computational resources than every baseline compared. SEdgeNet has the fewest parameters and the lowest multiply-accumulate count among the tested methods. Against DGCNN, the most efficient baseline, it uses 0.12 million fewer parameters and 0.02 giga-MACs less computation while achieving dramatically higher accuracy. Against AGConv, the most accurate baseline, it improves accuracy by 5.42 percent while cutting parameters by 0.52 million and computation by 8.43 giga-MACs. In controlled experiments swapping SEdgeConv for standard EdgeConv within an otherwise identical architecture, the stochastic version improved accuracy by 9.4 percent on MiliPoint and 0.44 percent on MMActivity, and required 3.08 giga-MACs less computation at a neighbourhood size of five. Ablation studies confirmed that both SNS and EFC contribute independently, and that combining them yields the best accuracy on both datasets alongside the lowest computational cost of 0.34 giga-MACs.

The hyperparameter analysis revealed instructive differences between the two datasets. MMActivity, with only five relatively separable classes, peaked with two layers and 35 nearest neighbours, reaching 98.12 percent accuracy, and its performance was dominated by network depth rather than neighbourhood size. MiliPoint, with 49 harder-to-distinguish classes, required a larger neighbourhood of 50 neighbours and five layers to reach 85.70 percent, suggesting that harder classification tasks demand both wider receptive fields and deeper networks to extract sufficiently discriminative geometric structure. The optimal sampling ratio also differed, at 0.6 for MMActivity and 0.3 for MiliPoint, though the authors found that a range of 0.3 to 0.5 delivers near-optimal accuracy on both datasets and recommend it as a practical default. Because the stochastic sampling remains active at inference time, the team also examined a Monte Carlo-style deployment strategy: averaging predictions over multiple forward passes. On MMActivity, accuracy rose from 97.72 percent with a single pass to a peak of 98.34 percent at six passes, at a still-modest cost of 2.04 giga-MACs.

Perhaps most compelling is the demonstration that the system runs in real time on embedded hardware. The researchers deployed their trained pipeline on a Jetson Nano mounted on a Robotino robot, with an AWR1843BOOST radar feeding data over USB. In a proof-of-concept trial, a single subject performed sitting, standing and walking freely in an indoor room, and the system achieved 99 percent classification accuracy, recognising actions under both lit and dark conditions. Average preprocessing latency was 12.18 milliseconds and model inference averaged 15.14 milliseconds, for a combined end-to-end latency of 27.32 milliseconds, comfortably within real-time requirements. The authors are careful to frame this as a feasibility demonstration rather than a comprehensive deployment study, noting that broader evaluation across multiple subjects, more activities and varying distances remains future work. Even so, the combination of privacy preservation, darkness-robust sensing, state-of-the-art accuracy and embedded real-time performance positions SEdgeNet as a promising foundation for assistive robotics, ambient intelligence and independent-living support, where the goal is to watch over people’s wellbeing without ever truly watching them.

Subject of Research: Graph neural network for human activity recognition from sparse millimetre-wave radar point clouds

Article Title: SEdgeNet: stochastic edge network for human activity recognition using sparse point cloud

Article References: Zakka, V. G., Manso, L. J., & Dai, Z. (2026). SEdgeNet: stochastic edge network for human activity recognition using sparse point cloud. Journal of Ambient Intelligence and Humanized Computing. https://doi.org/10.1007/s12652-026-05133-z

Image Credits: AI Generated

DOI: 10.1007/s12652-026-05133-z

Keywords: SEdgeNet, millimetre-wave radar, human activity recognition, point cloud, graph neural network, stochastic neighbour sampling, edge convolution, privacy-preserving sensing, embedded AI, ambient intelligence, depthwise separable convolution, assistive robotics

The primary challenge that this research addresses is the complex coordination required for a robot to use both of its arms in unison. The research group utilizes their advanced prototype, ADAM (Autonomous Domestic Ambidextrous Manipulator), which is specifically designed to perform a variety of assistive tasks in domestic settings. Unlike traditional robots limited to single-arm movements or reliant on pre-programmed routines, ADAM can autonomously carry out tasks such as serving meals and tidying spaces. This proactive approach to domestic assistance aims to support elderly individuals and those requiring additional help, allowing for safer and more comfortable living conditions.

ADAM is equipped with sophisticated features that enable it to analyze and adapt within various home scenarios. Researcher Alicia Mora, from the Mobile Robots Group at UC3M, emphasizes the significance of this flexible, adaptive learning style. For example, the robot can autonomously serve a glass of water or retrieve necessary items for an individual – tasks that, while seemingly simple to humans, encompass a myriad of complex movements and interactions challenging for robotic systems. Through the advanced capabilities of ADAM, the team envisions a future where robots can provide essential aid in the lives of those who need it most.

The technology underlying this robotic advancement hinges on a novel technique that allows each arm of the robot to learn tasks independently via imitation learning. Researchers Adrián Prados and Gonzalo Espinoza propose innovative methods whereby each arm learns to perform tasks in isolation before enabling seamless communication between them, utilizing a mathematical framework known as Gaussian Belief Propagation. This innovative technique simulates a constant, real-time dialogue between the robot’s limbs, allowing them to synchronize and navigate around obstacles or avoid collisions without interruptions for recalibrations. As a result, the movements executed by ADAM are not only more fluid and efficient but align closely with human-like interaction styles.

The core of this robotic learning paradigm centers around the idea that mere imitation is not enough. Current robotics technology often demands extensive programming and detailed code for even basic movements, whereas the UC3M approach focuses on enabling robots to learn through observation. By watching humans perform tasks, ADAM captures the intricacies of motion, empowering it to adapt to situations where exact replication would result in failure. Instead of rigidly repeating movements, the robot utilizes adaptive learning methods, making it capable of modifying actions in real-time, such as adjusting the trajectory of its limbs to compensate for unforeseen changes in the environment.

This adaptive learning resembles the properties of a “rubber band,” allowing the robot to preserve the essence of an action even when the context shifts. For example, if it learns to pick up a bottle, it can adjust its wrist and arm positioning to ensure the bottle remains upright, preventing spills, regardless of the original placement of the bottle. This cognitive flexibility not only increases efficiency but also enhances safety by ensuring that robotic assistance does not pose risks to its surroundings or the people it serves.

The functionality of ADAM is structured into three critical phases: perception, reasoning, and action. Initially, the robot perceives its environment by gathering data through an array of sensors. Advanced 2D and 3D laser sensors are utilized to measure distances, identify obstacles, and locate relevant objects within the space. Following perception, the reasoning phase involves processing the collected data to derive important insights and relevant information essential for decision-making. Finally, the action phase is where ADAM computes its movements, determining how best to execute tasks and coordinate its dual arms effectively.

One of the most challenging aspects of robotic implementation is transitioning from a mere visual analysis of objects to a deeper understanding of their usage and context. Conventional robotic systems relied on pre-programmed ‘common sense’ databases. However, UC3M researchers are now embedding generative models along with artificial intelligence into their system, enhancing ADAM’s capability to adapt to the specific requirements of real-world scenarios and the dynamic nature of human behavior.

While ADAM currently exists as an experimental platform with a development cost in the range of 80,000 to 100,000 euros, the technology demonstrates the potential for future widespread deployment in domestic environments, projected to become more economically accessible within the next 10 to 15 years as advancements continue. The implications of this research stretch beyond technical achievements and highlight the burgeoning role of robotics in addressing societal challenges, such as an aging population increasingly needing assistance in daily activities.

As global demographics shift, with an ever-growing number of elderly individuals and a declining pool of caretakers, solutions such as ADAM will play an integral part in addressing these societal needs. The vision shared by Ramón Barber, the director of the Mobile Robots Group at UC3M, is that robotic assistance will evolve into a critical resource for enhancing the quality of life and independence for vulnerable populations, thereby allowing individuals to maintain autonomy within their daily lives.

Thus, robotics is not simply about developing machines that can perform tasks—it represents a significant step toward enabling a future where technology complements human capabilities, enhancing overall well-being and quality of life. These innovative technologies, as highlighted by the research team’s dedication and methodology, signify a promising advancement toward a nuanced synergy between humans and machines that can transform societal living conditions and address essential needs.

As the field of robotic learning continues to advance, the ongoing research will undoubtedly inspire further innovation, opening doors for future collaboration between humans and intelligent systems. With every milestone achieved, we move closer to realizing the dream of seamlessly integrated robotic helpers that can intelligently and adaptively navigate the complexities of human environments.

In conclusion, the work being done at UC3M marks a revolution in the application of robotics for domestic assistance. This ambitious and thoughtful approach aligns with the evolving landscape of societal needs, showcasing how technological advancements can be harnessed to support a more independent and dignified life for individuals requiring assistance. The promising results of the ADAM project underline a future where robots are not merely machines but companions, co-workers, and invaluable contributors to enhancing daily life.

Subject of Research: Coordination of robotic arm movements using observational learning
Article Title: Coordination of Learned Decoupled Dual-Arm Tasks through Gaussian Belief Propagation
News Publication Date: 30-Nov-2025
Web References: [Insert relevant URLs]
References: [Insert relevant citations]
Image Credits: UC3M