Bipedal robots have achieved remarkable locomotion capabilities through reinforcement learning (RL), yet their real-world deployment remains hindered by the sim-to-real gap—dynamics mismatches between simulation and reality that degrade locomotion performance through behavioral deviations. This work introduces an online behavior adaptation framework that bridges this gap at the behavioral level by dynamically aligning emergent locomotion strategies with simulation-derived objectives. Our method integrates two core innovations: (1) a structured latent space constructed via an augmented Variational Autoencoder (VAE), which quantifies behavioral divergence through domain-invariant representations of locomotion patterns, and (2) a closed-loop adaptation module that maps latent-space deviations to real-time adjustments in low-level controller parameters. By reformulating sim-to-real transfer as a problem of behavioral alignment rather than explicit dynamics matching, the framework enables continuous adaptation to unmodeled dynamics mismatch without requiring system identification or offline retraining. Extensive experimental evaluations demonstrate the effectiveness of the proposed method, highlighting its potential to bridge the behavior gap between simulation and reality.

Bipedal robots are increasingly deployed in tasks where stability under load is critical. However, control methods relying solely on state feedback struggle to maintain stability under unknown foreign loads, as dynamics induced by the load directly affect balance critical states in intrinsically unstable bipedal systems. This paper proposes a foreign-load adaptive balance control framework for bipedal locomotion based on online load parameter estimation and composite model update. Load dynamic parameters are estimated online using an optimal method and embedded into the controller’s internal “robot with load” composite model to continuously update balance-relevant dynamics. Based on this representation, a hierarchical hip-ankle coordination strategy is adopted, where the hip provides proactive load compensation and the ankle regulates reactive posture stabilization and ground compliance. Simulation and hardware experiments demonstrate stable dynamic bipedal locomotion under heavy foreign loads with unknown dynamic parameters, validating improved balance robustness, load adaptability, and posture regulation for real-world applications.

In intraocular microsurgery with minute operational scales, instruments pass through non-visible regions of the anterior segment, where robot-assisted surgery, which heavily relies on visual perception, fails to determine the instrument’s attitude relative to the eyeball. This compromises surgical flexibility, increases risks, and hinders autonomous surgery development. Therefore, a framework for predicting instrument trajectories in non-visible regions during robot-assisted microsurgery has been proposed to mitigate the risks of retinal and lens injuries caused by blind operations and enhance surgical procedures’ intelligence and autonomy. First, a lightweight reconstruction of the anterior segment environment is performed under controlled knowledge guidance to construct a global map. Second, the tip position of the surgical instrument is detected through multi-sensor fusion, enabling the perception of instrument-environment interactions under visual constraints. Based on this, a long short-term spatiotemporal aggregation algorithm for instrument trajectory prediction is proposed, which enhances surgical safety by providing high-precision predictions of the instrument tip’s motion trajectory. Experiments show that the framework achieved a 0.0435 mm average prediction error in non-visible regions, corresponding to 0.03% of the region in a single dimension and 7.25% of the surgical instrument’s diameter. This significantly enhances the precision of robot-assisted surgery under visual constraints and provides robust technical support for safe, intelligent, and autonomous intraocular robotic surgery.

Online gait planning plays a crucial role for the locomotion of humanoid robots. While simplified models often fail to capture critical dynamic features of the robot’s motion, making online gait modifications with constrained nonlinear optimization in complex models is highly challenging with current computational power. This paper introduces a gait planning method that leverages supervised learning to expedite the gait planning and optimization process. Building upon our previous work, this paper extends a three-body model to the three-dimensional (3D) case. This model incorporates the angular momentum and height variation of body as well as the influence of leg motions, thus facilitating the generation of omnidirectional walking and running patterns. Due to the complexity of the model, an online gait planning modification is impractical. Therefore, supervised-learning is employed to train a policy derived from the model-based gait planning approach. This policy is then implemented online to produce versatile locomotion. Furthermore, the gradient of the trained neural network is utilized for nonlinear optimization of gait parameters, significantly improving the robot’s balance against external perturbations. The effectiveness of the proposed method is validated through a series of experiments conducted in simulation and on the real robot BHR-T, confirming its capability to generate adaptive walking and running motions in response to varying demands and disturbances.

Bipedal-vehicle transformable robots (BVTRs), equipped with driving wheels, combine the flexibility of bipedal locomotion with the speed of wheeled movement. However, maintaining balance across different formations under various external disturbances remains a significant challenge due to uncertain disturbance types and dynamic shifts between formations. To address these challenges, this paper introduces the concept of Equivalent Predictive Control (EPC), which models all disturbances as unified virtual wrenches and integrates them directly into the robot’s predictive control model, treated as an inertia-varying single rigid body. By anticipating the future impact of disturbances, EPC enhances stability and enables simultaneous handling of various disturbances. To address the challenge of dynamic changes, contact variations, and shifting constraints during formation transitions, we propose Handle Point Control (HPC). HPC simplifies multi-task tracking by reducing joint space control to a set of virtual target points, called ‘handle points’, such as knees, hips, and shoulders. This method facilitates real-time formation switching by tracking different handle points. Experiments on the BVTR platform BHR8-2 validate the effectiveness of the proposed control strategies. Note to Practitioners—This paper addresses two critical challenges for applying BVTRs in real-world industrial scenarios: 1) managing various external disturbances, and 2) overcoming the complexities associated with changing dynamics and contact situations during formation transitions. The proposed EPC strategy unifies all disturbance types and integrates them into the robot’s dynamic model, enhancing stability and adaptability across formations. The HPC method simplifies control by focusing on tracking key handle points, allowing smooth, real-time formation transitions without needing multiple optimization schemes. These methods can also be applied to other robotic systems that face similar control challenges.

This paper proposes a novel nonlinear model predictive control (NMPC) method based on geometric variational calculus for high-dynamic and complex motion control of quadruped robots. By approximating system trajectory tracking error dynamics on the Special Euclidean group (SE(3)), the method avoids the singularities of Euler angles and the challenges of quaternion representation while capturing the coupling between rotational and translational dynamics for a more comprehensive motion description. Leveraging variational calculus, the resulting Geometric Nonlinear Model Predictive Controller (GNMPC) enables high-frequency updates while preserving essential nonlinear system characteristics. Experimental results across various scenarios validate the effectiveness and advantages of the proposed controller. Note to Practitioners—The primary motivation of this paper is to investigate the application of geometric methods in Model Predictive Control (MPC) and to validate their effectiveness in the context of quadruped robots, which exhibit nonlinear dynamics. In this work, the authors model the robot’s motion on a nonlinear manifold and linearize the system using variational methods. Sequential Quadratic Programming (SQP) is then applied to approximate the globally optimal solution. Experimental results demonstrate that this approach significantly improves the performance of quadruped robots, particularly in handling highly dynamic and robust motions.

Biped robots have received increasing attention due to their human-like mechanical structure and good environmental adaptability. In this paper, a new Human-Simulated Intelligent Walking Control (HIWC) scheme is proposed to solve the stability problem of the most popular proportional differential (PD) control under model inaccuracy and disturbance, and further improve its control performance. Specifically, based on Human-Simulated Intelligent Control (HSIC), HIWC is a hierarchical control structure composed of a foot placement compensation (FPC) strategy at the high-level planning layer, and a multi-mode compensation controller (MCC) at the low-level (execution) layer. In FPC, a foot placement compensation algorithm is proposed to plan and correct the swing foots trajectory in real time. MCC consists of a PD and two adaptive compensation algorithms. MCC under bounded uncertainty is proven to be stable in this paper using the Lyapunov theorem. HIWC was tested and compared with PD and model predictive control (MPC) in three experiments on a physical robot platform for planar walking, push-pull, and uneven-ground walking. Experimental results show that the proposed HIWC is more flexible and accurate in controlling the robot’s movement.Note to Practitioners—This paper builds on the fact that PD controllers cannot be easily proven to be stable and do not provide accurate control for the biped robot walking problem. To address these issues, this paper proposes a novel control scheme namely Human-Simulated Intelligent Walking Control (HIWC) and belonging to the family of Human-Simulated Intelligent Control (HSIC) schemes. The proposed HIWC system has been compared against the model predictive control (MPC) and a proportional differential (PD) controller. The proposed HIWC, unlike PD controllers, is rigorously proven to be stable in the presence of model inaccuracy and disturbance. Furthermore, the experiments carried out on a real-world biped robot demonstrate the superiority of HIWC over PD and MPC in terms of control performance.

High dynamic jump motions are challenging tasks for humanoid robots to achieve environment adaptation and obstacle crossing. The trajectory optimization is a practical method to achieve high-dynamic and explosive jumping. This paper proposes a 3-step trajectory optimization framework for generating a jump motion for a humanoid robot. To improve iteration speed and achieve ideal performance, the framework comprises three sub-optimizations. The first optimization in-corporates momentum, inertia, and center of pressure (CoP), treating the robot as a static reaction momentum pendulum (SRMP) model to generate corresponding trajectories. The second optimization maps these trajectories to joint space using effective Quadratic Programming (QP) solvers. Finally, the third optimization generates whole-body joint trajectories utilizing trajectories generated by previous parts. With the combined consideration of momentum and inertia, the robot achieves agile forward jump motions. A simulation and experiments (Fig. 1) of forward jump with a distance of 1.0 m and 0.5 m height are presented in this paper, validating the applicability of the proposed framework.

Dynamic objects pose significant challenges to the accuracy of state estimation and map quality in Simultaneous Localization and Mapping (SLAM). While current dynamic SLAM methods often rely on semantic information to detect specific movable objects, this dependency on pre-trained models and semantic priors can lead to false dynamic detections. This paper presents a novel semantic-independent dynamic SLAM method that detects truly moving regions, without being constrained by the classes or motion patterns of dynamic objects. We introduce a geometric re-clustering approach to improve object clustering by addressing the under- and over-segmentation caused by the K-Means algorithm. Next, instead of simply classifying entire clusters as dynamic or static, we propose a method to detect dynamic regions within each cluster based on dense optical flow residuals. This enables the detection of partial object movements, such as a seated person moving only his hands. Dynamic detection results are propagated across consecutive frames as dynamic priors for calculating optical flow residuals. Additionally, to enhance map quality, we address the mis-detection of slowly or intermittently moving objects through depth consistency checks applied over a larger time interval. Extensive evaluations on public datasets (TUM and Bonn) and real-world scenes show that our method outperforms state-of-the-art semantic-based methods in terms of localization accuracy and generalizability across various scenarios, particularly when facing unknown dynamic objects. Our method also achieves clean and dense reconstructions, demonstrating its potential for applications like robot navigation in dynamic environments.

Internal limiting membrane (ILM) peeling is a vital vitreoretinal surgery procedure. However, due to the thickness of just 1-2 micrometers and the intricacies associated with its varying density and adhesion, the difficulty of manipulation exceeds the physiological limits of human perception and operation. Surgical robot is characterized by high precision and stability. However, navigating intricate intraocular environments and handling minuscule high-precision areas remain enormous challenges. These include issues of uneven lighting, field-of-view loss, and motion blur. This paper proposed a perception method named ‘Multimodal Surgical Process Recognition based on Domain Knowledge and Segmentation (MSPR-DKS),’ designed to address these challenges and provide input for the precise control of robots. Moreover, a comprehensive dataset focused on ILM peeling during macular hole surgeries was established. Experimental results underscore the efficacy of this approach, with segmentation accuracies exceeding 99.27% for instruments and macular holes and an average accuracy of 98.97% in recognizing surgical processes. This study paves the way for leveraging domain knowledge and image segmentation to improve robot-assisted manipulation of soft tissues in ophthalmology.

High-frequency and accurate state estimation is crucial for biped robots. This paper presents a tightly-coupled LiDAR-Inertial-Kinematic Odometry (LIKO) for biped robot state estimation based on an iterated extended Kalman filter. Beyond state estimation, the foot contact position is also modeled and estimated. This allows for both position and velocity updates from kinematic measurement. Additionally, the use of kinematic measurement results in an increased output state frequency of about 1kHz. This ensures temporal continuity of the estimated state and makes it practical for control purposes of biped robots. We also announce a biped robot dataset consisting of LiDAR, inertial measurement unit (IMU), joint encoders, force/torque (F/T) sensors, and motion capture ground truth to evaluate the proposed method. The dataset is collected during robot locomotion, and our approach reached the best quantitative result among other LIO-based methods and biped robot state estimation algorithms. The dataset and source code will be available at https://github.com/Mr-Zqr/LIKO.

Recent advances in both legged robot locomotion and Reinforcement Learning have shown a promising path for developing bipedal robot controllers. While the difference in dynamics between real world and simulation, also known as reality gap, still hinders the use. In this paper, we focus on sim-to-real bipedal robot locomotion task. We leverage the recent advances in auto-tuning sim-to-real transfer and use it to address sim-to-real bipedal robot locomotion problem. Similar to existing work, we first train a parameter searching model with dataset collected from simulator and use real-world data to tune the simulation parameters. However, the prediction tuning can be unreliable if the training dataset distribution fails to cover the real-world data. We address this problem by formulating this problem as an Out-of-distribution problem and further extending the current framework with a dataset verification model. With extended module, our method is capable of tuning the simulation parameters safely and efficiently. We demonstrate our method outperforms existing work and achieves sim-to-real bipedal robot locomotion on bipedal robot BITeno.

Feasible regions for continuous walking must provide necessary information for footstep planning, including surrounding landing areas and details about obstacles to be avoided during foot swing. However, the current frame lacks sufficient information to construct a feasible region needed at the current moment due to knee occlusion. To this end, this paper uses polygon merging to construct an information-complete feasible region. This polygon merging refers to merging polygons from the current frame and a specific previous frame. Since the polygon is more concise and efficient than point cloud for environmental representation, construction can be completed quickly without GPU acceleration. Experiments show that the proposed method successfully constructs informative feasible regions within the allowed time frame, enabling the robot to navigate stairs.

The optimization of robot controller parameters is a crucial task for enhancing robot performance, yet it often presents challenges due to the complexity of multi-objective, multi-dimensional multi-parameter optimization. This paper introduces a novel approach aimed at efficiently optimizing robot controller parameters to enhance its motion performance. While spiking neural P systems have shown great potential in addressing optimization problems, there has been limited research and validation concerning their application in continuous numerical, multi-objective, and multi-dimensional multi-parameter contexts. To address this research gap, our paper proposes the Entropy-Weighted Numerical Gradient Optimization Spiking Neural P System, which combines the strengths of entropy weighting and spiking neural P systems. First, the introduction of entropy weighting eliminates the subjectivity of weight selection, enhancing the objectivity and reproducibility of the optimization process. Second, our approach employs parallel gradient descent to achieve efficient multi-dimensional multi-parameter optimization searches. In conclusion, validation results on a biped robot simulation model show that our method markedly enhances walking performance compared to traditional approaches and other optimization algorithms. We achieved a velocity mean absolute error at least 35% lower than other methods, with a displacement error two orders of magnitude smaller. This research provides an effective new avenue for performance optimization in the field of robotics.

For humanoid robots, online motion generation on non-flat terrain remains an ongoing research challenge. Computational complexity is one of the primary restrictions that preclude motion planners from generating adaptive behaviors online. In this paper, we investigate this problem and decompose it into two sequential components: an Efficient Behavior Generator (EBG) and a Nonlinear Centroidal Model Predictive Controller (NC-MPC). The EBG is responsible for optimizing the physically feasible whole-body template behaviors, which can provide reliable warm-starts for NC-MPC, thereby greatly reducing the computational effort of online planning. With tailored objective function and feet complementary constraints, the EBG can search for a near-optimal solution after several iterations within seconds for different behaviors including walking, running, and jumping, even with intuitive initial guesses. To make the template behaviors extensible when the robot encounters possible different scenarios, the NC-MPC is proposed to regenerate the reactive motion online to adapt it to the real local environment. Finally, we validate the effectiveness of synthesizing EBG and NC-MPC for humanoid locomotion on non-flat terrain in simulation and on the real humanoid robot BHR7P.Note to Practitioners— For current humanoid robots, dynamically traversing non-flat terrain such as stairs, slopes, and gaps in the real world presents a significant challenge. In this paper, we propose an adaptive motion planner for humanoid robots to traverse non-flat terrain, which is properly integrated into the closed loop of online control. Considering computational complexity and motion extensibility, the planner consists of two parts: an efficient behavior generator performed offline and a nonlinear model predictive controller performed online. The behavior generator can efficiently generate template behaviors for the humanoid robot, including various gaits such as walking, running, and jumping. To make these template behaviors adaptable, a nonlinear model predictive controller based on the centroidal dynamics model is developed to plan reactive motions online. It can extend template behaviors to fit potentially different scenarios in practice. The proposed method is validated in simulations and experiments with the humanoid robot BHR7P. Furthermore, this method can be applied to legged robots or systems that need to move dynamically on non-flat terrain, such as quadruped and hexapod robots.

Highly dynamic movements such as jumping are important to improve the agility and environmental adaptation of humanoid robots. This article proposes an online optimization method to realize a vertical jump with centroidal angular momentum (CAM) control and landing impact absorption for a humanoid robot. First, the robot's center of mass (CoM) trajectory is generated by nonlinear optimization. Then, a quasi-sliding mode controller is designed to ensure that the robot tracks the CoM trajectory accurately. To avoid unexpected spinning in the flight phase, a center-of-pressure-guided angular momentum controller is designed to stabilize the CAM. The modifications of CoM and CAM are realized by online optimization of dynamic components and inverse dynamics. Two quadratic programming optimizations are utilized to generate feasible contact force/torque and joint acceleration referring to uplevel CoM and CAM controllers. In addition, a viscoelastic model-based controller is designed to absorb the vibration caused by a large contact impact. A simulation and experiment of a 0.5-m high (foot lifting distance) vertical jump are achieved on a humanoid robot platform in this article (Fig. 1).

Stability control for humanoid robots based on zero moment point (ZMP) control and impedance control are widespread. However, uncertain changes in the center of mass (CoM) height for ZMP control and specific regulation of the variable stiffness of impedance control have been challenging issues in previous studies. In this article, these two problems are solved by implementing fuzzy control-based regulations. First, the fuzzy ZMP controller, which regulates the feedback gains online based on the CoM height change and CoM tracking errors, is proposed. Second, we propose a fuzzy regulation law for variable stiffness, which is applied for uncertain contact situations and inspired by the pattern of human muscle stiffness. With these two methods, the ground adaptability for humanoid robots is enhanced. The proposed method is validated with experiments on a real robot platform, BHR-T.

Compliance is important for humanoid robots, especially a position-controlled one, to perform tasks in complicated environments where unexpected or sudden contacts will result in large impacts which may cause instability or destroy the hardware of robots. This paper presents a compliance control method based on viscoelastic model for humanoid robots to survive on these conditions. The viscoelastic model is used to obtain the relationship between the differential of contact force/torque and linear/angular position. Thus a state equation of this model can be established and a state feedback controller adjusting the position to adapt to the contact force/torque can be designed to realize the compliant movement. The proposed compliance control method based on viscoelastic model has been employed in ankle compliance for stable walking on indefinite uneven terrain and arm compliance for falling protection on BHR-6P, a position-controlled humanoid robot, which validates its effectiveness.

Generally, large knee joint torque is required when a leg is flexed. However, the large torque motor will increase the robot size and total weight. Also, a large reduction ratio gear box to realize the large torque will decrease output speed and lower backdrivability of the joint. It makes the robot difficult to perform agile and flexible motion like animals. In this paper, we propose a robot leg with a knee joint mechanism consisting of a variable reduction ratio crossed four-bar linkage mechanism (VRRCFLM) based on cruciate ligament of an animal. The VRRCFLM is to increase the reduction ratio for large knee flexion postures without greatly reducing the total backdrivability. In this paper, we developed a robot leg with a knee joint mechanism consisting of the VRRCFLM. In order to design the link parameters of the leg mechanism, optimization design aimed at maximizing the jumping height of the robot was performed. The robot model with the designed mechanism was evaluated through the dynamics simulations. Thanks to the VRRCFLM, the required torque of the knee motor at the large flexion postures was decreased. Moreover, the vertical jumping height was improved by 24.6 % comparing with a model without the mechanism. In experiments using the prototype, the required static torque was decreased as in simulation, and the jumping height was more than one leg length.

Humanoid robots are being designed to perform tasks currently carried out by human workers in industry, manufacturing, service, and disaster assistance. To this end, the humanoid robot should be able to walk stably across many types of terrain. However, when traversing a complex unknown environment, it is difficult to realize accurate terrain perception immediately through large data collected by the sensor system, leading to a difference between planned foot landing positions and actual foot landing positions. As a result, an unexpected contact force/torque may affect the stability of the robot. This paper adopts active contact perception instead of terrain perception and proposes a contact force/torque control method based on the viscoelastic model to address this problem. In addition, we design a body stability controller based on tracking the trajectories of the virtual repellent point (VRP) and the divergent component of motion (DCM) to restrain the disturbance caused by the unexpected contact force/torque. Simulations and experiments on the BHR-6P humanoid robot platform demonstrate the proposed contact force/torque control method for walking on indefinite uneven terrain.

For real-world applications, a biped robot should maintain stable walking when subjected to sudden external disturbances. Typically, unexpected changes to body acceleration indicate that a robot is experiencing an external disturbance. This paper presents a biped walking controller for rapid response to large external disturbances. First, a novel adjustment algorithm for foot placement is proposed. Here, zero-moment point variations are mapped onto the new foothold based on calculations from changes in body acceleration. Second, a novel impact reduction control for foot landing is presented based on abating body vibrations. To avoid false detection triggers and excessive foothold adjustment, a rapid disturbance detection method is established using the body acceleration derivative. Finally, the effectiveness of the proposed methods is validated under simulations and in experiments with an actual biped robot.

Cats have protractible claws to fold their tips to keep them sharp. They protract claws while hunting and pawing on slippery surfaces. Protracted claws by tendons and muscles of toes can help cats anchoring themselves steady while their locomotion trends to slip and releasing the hold while they retract claws intentionally. This research proposes a kind of modularized self-adaptive toe mechanism inspired by cat claws to improve the extremities' contact performance for legged robot. The mechanism is constructed with four-bar linkage actuated by contact reaction force and retracted by applied spring tension. A feasible mechanical design based on several essential parameters is introduced and an integrated Sole-Toe prototype is built for experimental evaluation. Mechanical self-adaption and actual contact performance on specific surface have been evaluated respectively on a biped walking platform and a bench-top mechanical testing.

When a biped robot moves about in a physical environment, it may encounter inclined ground. Biped walking on inclined ground still remains challenging for biped robots. Previous studies have discussed biped walking on inclined ground along specific directions. However, omnidirectional walk on inclined ground has rarely been investigated. In this paper, we propose a gait pattern generation method for omnidirectional biped walking on inclined ground. First, a model that describes the motion of biped walking on inclined ground uniformly with two angle parameters is proposed. A mathematical relationship between motions in the sagittal and coronal planes of the biped robot are presented. Then, based on nonorthogonal motion decoupling, a method that generates gait patterns for omnidirectional walking with a double support phase for biped robots is proposed. The trajectories of each foot are designated by the walking speed, step length, and walking direction. The motion trajectory of the center of mass (CoM) of the robot is planned using a linear inverted pendulum model in the sagittal and coronal planes. The motion of CoM in the sagittal and coronal planes is constrained in parallel to the gradient vector of the inclined ground and the horizontal plane, respectively. Finally, the effectiveness of the proposed gait planning method for biped walking on is validated by simulations and experiments with an actual biped robot.

To enhance the safety of a humanoid robot when it is operating a complex environment, a number of methods that combine visual and force information have been presented. These methods are generally divided into two approaches. The first approach is to coordinate the visual controller and force controller in a parallel way, and the second approach is to coordinate them in series. However, these two approaches do not consider the interaction between the visual controller and force controller. Specifically, the first approach does not consider the interaction between the controllers. The second approach only considers the effect of the output of the visual controller on the force controller, while the effect of the force controller on the visual controller is not considered. This study presents a design for a new flexible controller for a humanoid robot that considers the interaction of visual and force information. The advantages of the proposed method are that it simultaneously incorporates the functions of a visual servo controller and a flexible controller as well as its ability to consider the interaction of visual and force information when a humanoid robot is operating.

An improved ZMP (Zero Moment Point) trajectory for a biped robot is designed in this paper, which imitates a human's actual ZMP trajectory in the walking process. A new method of walking pattern generation based on forward moving ZMP in SSP (Single Support Phase) is also provided. It can keep the ZMP moving forward instead of staying in the center of supporting region during SSP, which is helpful for increasing the walking speed. We have been developing BHR, which has 38 DOFs (degree of freedom). The effectiveness of the method is conducted by simulation and walking experiment on BHR.