Automatic video-based road inspection provides a promising method for evaluating the visibility and legibility of traffic signs. However, detection reliability deteriorates due to fluctuations in image quality caused by long-range and oblique-angle shooting. We propose a detection framework for feature-deficient inspection imagery. The framework jointly learns local descriptors and global context, enhancing fine-grained discriminability while maintaining semantic integrity. Building on the YOLOX architecture, we introduce an optimized detection algorithm, YOLO-TRs, which leverages residual skip aggregation to preserve high-frequency spatial details in deeper layers. In the neck, we capture cross-scale dependencies through the adaptive integration of multi-dimensional spatial context and reweight adjacent pyramid levels using learned affinities to promote cross-layer information interaction. These mechanisms reinforce discriminative feature representations under feature-deficient imaging. We construct a purpose-built traffic sign database, TSDB-24, that emulates the target semantic features encountered during road inspections by continuously controlling pixel coverage and deformation rate. By testing images with varying characteristic parameters, we quantify their effects on detection accuracy and robustness, and further conduct a failure modes and effects analysis to trace the causes of algorithmic failures to specific image factors. Comparative experiments on the TSDB-24 and the public CTSDB databases demonstrate that our method improves mAP over the baseline by 7.48%. Under low-pixel-coverage and high-deformation regimes, our method achieves gains of 13.65% and 8.41%, respectively, outperforming prominent algorithms in accuracy and robustness. It is hoped that safety-critical traffic management can be supported by applying the developed method to road inspections under real-world conditions.

How to optimally synchronize exoskeleton powered assistance remains a problem that limits the broad application of such devices. Kinematic change during frequent switching between go and stop, a common and representative activity of daily living (ADL), makes it challenging to predict the gait phase and deliver assistance due to unpredictable movements. Conventional adaptive oscillators (AO) have been verified to be effective in gait phase prediction in steady-state walking. When walking cadences are changed, it usually requires multiple walking strides to synchronize the assistance, causing inaccurate or even unwanted force disturbances that can be dangerous in some cases. To solve this problem, a gait-based AO is proposed in this paper. It has two-level AO systems. The high-level AO learns from last stride and updates the low-level AO, which is designed to estimate the gait phase of the current stride in real time. This approach significantly decreases the time required to learn walking kinematics, while simultaneously improving the accuracy of predictions. Experiments were conducted on seven participants, and the results showed that the proposed gait-based AO can predict the gait phase faster, more accurate, and more stable than the conventional one during non-steady-state walking. Note to Practitioners—This article developed a new gait phase prediction method called gait-based adaptive oscillator. It aims to provide fast and reliable gait phase prediction in situations characterized by frequent transitioning between stop and go (non-steady-state walking). This method has the potential to be an alternative to existing phase prediction methods for robotic exoskeleton assisted walking in daily life, as it requires no model training before use while can synchronize exoskeleton assistance within two strides.

Multiple unmanned aerial vehicles (UAVs) are able to efficiently accomplish a variety of tasks in complex scenarios. However, developing a collision-avoiding flocking policy for multiple fixed-wing UAVs is still challenging, especially in obstacle-cluttered environments. In this article, we propose a novel curriculum-based multiagent deep reinforcement learning (MADRL) approach called task-specific curriculum-based MADRL (TSCAL) to learn the decentralized flocking with obstacle avoidance policy for multiple fixed-wing UAVs. The core idea is to decompose the collision-avoiding flocking task into multiple subtasks and progressively increase the number of subtasks to be solved in a staged manner. Meanwhile, TSCAL iteratively alternates between the procedures of online learning and offline transfer. For online learning, we propose a hierarchical recurrent attention multiagent actor-critic (HRAMA) algorithm to learn the policies for the corresponding subtask(s) in each learning stage. For offline transfer, we develop two transfer mechanisms, i.e., model reload and buffer reuse, to transfer knowledge between two neighboring stages. A series of numerical simulations demonstrate the significant advantages of TSCAL in terms of policy optimality, sample efficiency, and learning stability. Finally, the high-fidelity hardware-in-the-loop (HITL) simulation is conducted to verify the adaptability of TSCAL. A video about the numerical and HITL simulations is available at https://youtu.be/R9yLJNYRIqY.

A swarm of fixed-wing unmanned aerial vehicles (UAVs) is expected to efficiently accomplish various tasks in complex scenarios. This paper proposes an attention-based population-invariant multi-agent deep reinforcement learning (MADRL) approach to deal with the decentralized collision-free flocking problem for a scalable fixed-wing UAV swarm. First, this problem is modeled as a decentralized partially observable Markov decision process from the perspective of each follower. Then, an improved multi-agent deep deterministic policy gradient (MADDPG) algorithm is presented to efficiently learn the population-invariant flocking policy. In this algorithm, the parameter sharing with ego-centric representation mechanism is incorporated to improve learning efficiency. Besides, the attention-based population-invariant network structure (APINet) is designed by leveraging the self-attention mechanism. With this structure, the learned flocking policy is invariant to the population of the swarm. Finally, both numerical and hardware-in-the-loop simulation results verify the efficiency and scalability of the proposed approach.

Drone has become more and more popular in various civil applications due to the open of the low-altitude airspace and its easy operation. Unlike the common search and track task for the target, the new search and lock (SAL) task is focused on in this article. In this task, multiple drones first try to detect the moving ground target cooperatively. Then, they must lock the target by covering all the surrounding area of it at a low flight altitude, which indicates that the target can be watched clearly in all directions from then on. The SAL task can be applied in the panorama shot for the moving ground target. First, the low-altitude urban airspace is discretized into cubes, based on which the flight rules of drone are defined. The field of view (FOV) of drone is modeled considering the flight altitude and the block of buildings. For the cooperation among multiple drones, the constraints on the patten of waypoint, the communication distance, and the collision avoidance are all included. The goal in the search phase is to cover more area, which have not been visited recently to increase the probability of detecting the target, and it is expected to lock the target as soon as possible in the lock phase. A new swarm-based imitative learning optimization (SBILO) algorithm is proposed to determine the waypoint of drone in the search phase. To have a quick response to the escape behavior of the target in the lock phase, the waypoint of drone is generated in a distributed way to cover more surrounding area of the target and lock it gradually. The case of losing the target in the FOV of all drones is also addressed by covering more possible places where the target may appear. Simulation results demonstrate that the SAL task can be performed efficiently by the drones with the flight routes obtained by the proposed SBILO algorithm and the distributed asynchronous decision-make (DADM) approach.

In this article, a constrained image-based visual servoing control method for the shipboard landing problem of unmanned helicopters is proposed. First, the pitch and roll motion of ship are predicted by an autoregressive (AR) model to determine an appropriate period for landing. Subsequently, a novel robust sliding mode controller without linear velocity measurements is developed on the basis of the perspective image feature in a virtual image plane. Meanwhile, a modified Chebyshev neural network (CNN) is proposed to estimate the uncertainties, including the linear acceleration of ship and translational perturbation, while an adaptive law is employed to compensate the influence of rotational disturbances. The whole controller only requires the measurements feedback of a vision sensor and an inertial measurement unit (IMU). Ulteriorly, to prevent the visual target on the ship from going beyond the field of view of camera, the constrained controller is developed by a control barrier function and a quadratic programming, where the unknown relative velocity is estimated by a velocity observer. Finally, simulations are implemented to substantiate the capability of the presented shipboard landing control method.

Path planning is of critical importance to Unmanned Aerial Vehicle (UAV) based applications like unmanned delivery and inspection. The operating space of UAVs in urban environments is usually complex and dynamic. Safety and efficiency are hence the major concerns for UAV operations. Traditional path planning algorithms tend to find trajectories for UAVs in terms of safety which are theoretically optimal. However, the objective is usually difficult to achieve in practice given the complex operating environments and limited UAV capabilities. Thus, a path planning algorithm for UAVs that finds feasible trajectories is needed for UAV-based operations. In this paper, an Artificial Potential Field (APF) based algorithm is proposed for UAV's trajectory planning considerings concerns of safety, efficiency and kinetic constraints of UAVs. Simulation results show an improvement of 23% on efficiency and 89.5% on smoothness of trajectories over existing heuristic based methods. Performance metrics of safety, efficiency and smoothness of trajectories are analyzed based on simulation results. Parameter values in obtaining the good performance metrics have been suggested in the end of the paper.

Adaptive grippers are often used to fulfil the task required of grasping or manipulation with a variety of geometry attributes. In this paper, we present the design and evaluation of underactuated adaptive fingers for the common parallel grippers. A design methodology is proposed to define contacts and grasping stability. A simplified adaptive finger has also been designed, assembled and tested. The developed adaptive finger is able to perform encompassing and pinch grasps. The experiments show that the fingers are able to form three to six contact points for encompassing grasps and two contact points for pinch grasps.

This paper presents a novel structure-controlled variable stiffness robotic gripper that enables adaptive gripping of soft and rigid objects with a wide range of compliance. With the structure-controllable principle, the stiffness is controlled by the mechanical structure configurations rather than by material properties or electronic means. The principle is realized by changing the effective second moment of area of the gripper finger through rotating a built-in flexure hinge shaft. Based on this principle, the states of the stiffness can be continuously, instead of discretely, studied and assessed over the intermediate states from compliant to almost completely rigid. A variable stiffness mechanism has been developed to demonstrate the validity of the proposed principle. It enables that the finger stiffness and gripping position are independently controlled. With the introduction of flexure hinges, the undesired lateral buckling resulted from the rotation of a normal leaf spring is eliminated. In addition, a two-finger parallel gripper with this variable stiffness mechanism is developed which can provide the grasping stiffness according to the grasping task requirements. The effectiveness of the gripper has been demonstrated to handle the objects range from light, fragile to heavy, rigid without using any feedback loop or soft pads.

Animals provide inspiration for developing soft machines in bionics, robotics research as well as potential applications. This paper presents an integrated development of locomotive soft robot platforms: starfish-inspired robots with multi-limb bodies actuated by shape memory alloys (SMAs). The designs of these robots were based on the biological specifications of the locomotion and water-vascular systems of regenerating starfish. Multi-limb robot prototypes were fabricated with soft materials and 3D printing technology, satisfying modest movement requirements with SMA spring actuators. Experimental testing of multi-limb robots were conducted in accordance with biological locomotion principles. The robots vividly displayed typical starfish motion models. Animals provide inspiration for developing soft machines in bionics, robotics research as well as potential applications.

Automating item picking in the e-commerce warehouse is pressing but challenging, due to a massive variety of items, tight environmental constraints and item location uncertainty. In this paper, we present an effective and efficient strategy-based planning approach to implement the robotic picking from shelves for e-commerce. Making full advantage of a gripper with multiple securing methods, differentiated strategies are modeled as picking primitives with different securing methods. A strategy generator is proposed to produce feasible potential pickings as quickly and as successfully as possible. A strategy evaluator considering reachability, collision, object-bias preference and the securing performance is also presented for ranking the picking strategies. Experiments were conducted to validate that the robotic picker is able to plan a picking strategy within 2 ms and pick daily items from the shelves with an average success rate of 68%.

This paper presents a new flexible fixtureless assembly (FFA) approach for building T-joint structures automatically. The method is based on a Fixturing Assistive Gripper (FAG) mounted on a robot arm for forming T-joint sections. It eliminates possible uncertainties in terms of positioning and size errors during fit-up process. The method is verified by a robotic assembly workcell which includes dual robots and a simple machine vision unit. The basic algorithms for image processing, task planning, dual robot cooperative motion and path planning are developed for assembling a simple structure which predominantly formed by T-joint sections. Experiments show that the workcell can effectively eliminate the alignment error and successfully complete T-joint assembly task without the complicated sensing system typically required in the existing approaches.

Robotic exoskeletons have drawn much attention recently due to their potential ability to help the stroke and spinal cord injury patients to regain the ability of walking. However, the biggest challenge is the balancing of the exoskeleton and how it can balance is still an open question. Most of the time, patients using such exoskeleton devices require sufficient upper body strength to control upright posture and also manipulate crutches/walking frames to partially support body weight and keep balance. However, high energy cost and the high potential of falling using these devices remains a problem. In this paper, the issues are tackled by virtue of a proposed balance stabilizer mechanism which is able to provide active balance assistance for robotic exoskeletons. The design of a robotic exoskeleton together with balance stabilizer mechanism will be presented and discussed. In addition, a trajectory generation method, which can generate dynamically stable and tunable gait pattern, will also be shown. Finally, clinical trial results with a tetraplegia subject is presented and discussed.

A two-dimensional perching model is proposed first for perching with quadrotors. Then a perching mechanism by means of grasping is designed based on the model. The kinematic specifications of the perching mechanism are optimized to maximize the force transfer ratio so that sufficient grasping force can be generated for reliable perching. A controller of the gripper based on the control strategy from previous development is designed for autonomous perching with a quadrotor. Experiments on the grasping capability and reliability of the mechanism and its effectiveness with the controller for autonomous perching are conducted. Results show that the perching mechanism can generate sufficient grasping force and achieve autonomous perching to a target pole with a quadrotor both effectively and reliably.

In this paper, biomimetic undulating fins are considered with the focus on their hysteresis nonlinearity. Hysteresis is confirmed with experimental data on the biorobotic fin prototype, and qualitative modeling on this nonlinear action is then achieved by using Preisach equations. The developed iterative learning control is applied to eliminate hysteresis nonlinearity in biorobotic undulating fins. Both the simulation and experimental results show that the proposed control method is effective and feasible to improve the tracking performance of biorobotic fins by considering the hysteresis effect. Furthermore, the control methods should facilitate biomimetic investigation on propulsive modes and waveforms of fish swimming.

A gait trainer, NaTUre-gaits (natural and tunable rehabilitation gait system), has been developed to provide assistance for the gait rehabilitation. The gait rehabilitation system can provide mobility for an overground locomotion in forward walking and turning by a mobile platform. The exoskeleton modules are mounted on the mobile platform and attached to the lower limbs and pelvis in parallel. The synchronized motion generation for the exoskeleton modules is provided by virtue of the inverse kinematic model analysis. The pelvic trajectory is predefined and ten points are specified within one gait cycle to obtain the foot trajectory from the designated step length and height. The trajectory is obtained via curve fitting based on these specified points. On the other hand, in order to keep the desired foot trajectory, the pelvic motion is compensated to accommodate hip and knee joint angles. The proposed generation method is tested on the gait system and it has demonstrated the effectiveness and smoothness of the system.

In this paper, a fish robot employing a carangiform swimming mode is adopted as an experimental platform for the parametric study. Experiments conducted in the laboratory aim to study the variation of robotic fish's thrust with respect to various parameters including the frequency and amplitude of oscillation, joint link, aspect ratio, free flow velocity and the spring effect. The testing also enables us to find out the relationship between various parameters and the thrust generated by the oscillatory motion of tail. On the other hand, the significance of the controlled parameters will be determined using statistical methods.

To develop useful applications for the underwater robots, the first step is to study its performance, such as thrust force or swimming velocity. In order to avoid numerous parameter studies in evaluating its performance, the prediction model plays an important role. As fish's swimming includes the kinematic of its own body and the hydrodynamic interaction with the surrounding fluid, it is difficult to formulate a precise mathematical model by purely analytical approaches. This paper offers a semi-empirical method to model the performance of a BCF (body and/or caudal fin) biomimetic fish robot. By using a dimensional statistical method, a semi-empirical model for predicting the thrust force generated by a BCF oscillation swimming mode is derived. This model shows good prediction capability. The predicted results are in good agreement with the experiment data. Therefore, the proposed modeling method can be used to solve the engineering problem concerned without a complex theory derivation.

Gait pattern planning is an important issue in robotic gait rehabilitation. Gait pattern is known to be related to gait parameters, such as cadence, stride length, and walking speed. Thus, prior before the discussion of gait pattern planning, the planning of gait parameters for natural walking should be addressed. This work utilizes multi-layer perceptron neural network (MLPNN) to predict natural gait parameters for a given subject. The inputs of the MLPNN are age, gender, body height, and body weight of the targeted subject. The MLPNN is trained to output a suitable walking speed and cadence for given subject. Two MLPNNs are trained to study the efficiency and accuracy in predicting the desired outputs, for two different setups. First setup is that the MLPNN is trained specifically for slow speed condition only. In second setup, the MLPNN is trained for both slow and normal speed conditions. The results of the MLPNNs are presented in this paper. The efficiency and accuracy of the MLPNNs are discussed.

In the present work we study the station-holding performance exhibited by real fish and implement the similar performance onto a biomimetic fish robot. The aim is to make the swimming of a fish robot capable of holding its station in one-dimensional unsteady adverse flow, i.e. to obtain adaptation to the variations of flow rate. With the sensory feedbacks including water environment information and the displacement of fish robot, a closed-loop swimming control scheme is developed. Firstly, inspired by a real carp fish swimming in adverse flow, control laws that regulate the tail beat frequency and amplitude of fish robots are formulated. A non-linear oscillator is applied to model the periodic swimming gaits. As controlling parameters of the oscillator, frequency and amplitude are dynamically tuned by the feedback control laws. The output of the oscillator that generates the adaptive swimming gaits is adjusted accordingly. The control method is verified through simulation and experiments on a BCF type fish robot.

The present work is motivated by the need to develop a generic method modeling the biomimetic undulatory motion for fish robots with long fin propulsors. Combined with mechanical design of long fins proposed in current literatures, we explore the application of coupled nonlinear oscillators in the modeling of swimming gaits and propose a kinematic modeling framework. Coupled nonlinear oscillators can also be regarded as models of artificial Central Pattern Generators (CPGs) for swimming gait control of fish robots. The advantages of this method over the normal sinusoidal functions based method are discussed. The synchronization of multiple oscillators is derived, which can be utilized for the coordination of multiple joints of fish robots and the online gait transition. The framework is applied and tested in swimming motion control of an eight-DOF undulatory fin prototype. The effectiveness of the control is shown through experiments.

This paper discusses the implementation of the biological fish swimming motion onto the biomimetic fish robots. The mechanism designs and the control methodologies might vary as the fundamental knowledge is learnt from different species of fish. However, all those systems can be divided into two major types in general, namely: serial open chain design and parallel mechanism design from an engineering viewpoint. The general solutions to the swimming motion planning for the type with parallel mechanism are discussed by using the kinematics equations, followed by the derivation of analytical gait control functions. The application of these solutions on a manta fish prototype and an eight-joint MPF (media and/or paired fin) fish prototype shows the good agreement with the proposed modeling in the steady swimming and forward/backward swimming.

The objective of this work is first to present a study and analysis of fundamental issues that constitute the development of a hand rehabilitation system by making use of mechanisms for objective and quantitative assessment for patients with motor weakness, such as spinal cord injury (SCI) and post-stroke. The proposed rehabilitation system aims to improve the functional ability of impaired hand on the identified functional activities of daily living (ADL) and rehabilitation exercises. The system, through an interactive control, assists the subject to undergo activities of daily living (ADL) task relevant training on the impaired hand. The work also contributes in closing the knowledge gap of clinical and engineering groups. A kinematics modeling for the human hand-finger model is derived. The joint model has then been linked to rehabilitation functional activities of daily living (ADL). According to the kinematics studied and the joint ranges identified, a single five-bar mechanism assisting the finger motion is proposed. The kinematics of the five-bar mechanism is derived and the inverse solutions are obtained. An 11-DOF load-free hand prototype formed and assembled by a set of the five-bar mechanisms is described with four associated modules. Finally, the preliminary works and future research are concluded in this paper.

This paper presents an environment-friendly robotic system mimicking undulating fins of fish. To mimic the actual flexible fin of a real fish, we model a fin-like mechanism with a series of connecting linkages and attached it to the robotic fish. By virtue of a specially designed strip, each link is able to turn and slide with respect to the adjacent link. The driving linkages are used to form a mechanical fin consisting of several fin segments, which are able to produce undulations, similar to those produced by the actual fins. Owing to the modular design and the flexible structure of the mechanical fin, we were able to construct biomimetic robot fish with various swimming modes by fin undulations. Some qualitative and workspace observations by experiments with the robotic fish are shown and discussed

In the basic impedance control structure, it is usually required to know the environment stiffness/impedance for precise force tracking. In addition, the performance may be degraded when the random external disturbances exist. By incorporating a PI-type compensator to the basic impedance control structure, this paper presents a combined impedance/direct control scheme for control of robot manipulators, which is characterized by high disturbance rejection, robustness to uncertainties, and simple structure for application. Moreover, it does not require to switch the control modes during the transition between the free space motion and the constrained motion. Therefore, the proposed controller combines the advantages of both the direct control and the impedance control. Computer simulations are performed to demonstrate the superiority of the proposed control scheme over the impedance control.

In this paper, a study of a wheeled mobile robot (WMR) with omni-directional mobility and rover-like climbing ability is presented. Various rovers with robust locomotion design had been developed. However, the rover locomotion design with omni-directional mobility has not been addressed commonly. Shrimp suspension system and parallel bogies were employed as robot locomotion. It adapts to the terrain profile passively and it has better climbing capability than other rovers. To achieve a singularityless motion, the kinematic model of a wheeled mobile robot (WMR), including a platform equipped with six omni-directional wheels (ODWs), is formulated. The contents of this study are the design principle, kinematics, and motion studies. Also, experimental results have been employed to verify the model developed.

An environment-friendly propulsion system mimicking undulating fins of cuttlefish or stingray has been built. A nonconventional method was considered to model the flexibility of the fins of cuttlefish or stingray. A two-degree-of-freedom linkage was designed and constructed to mimic the actual flexible fin. Several linkages were used to form a mechanical fin consisting of several fin segments, which are able to produce undulations, similar to those produced by the actual fins. A number of experiments were carried out on the mechanical fin attached to a buoyant tank, which mimics a fish. The results were found consistent with anguilliform swimming behavior.

This paper presents an option in solving the problem with micro-assembly, i.e. long travel range with very high accuracy demand. A flexure-based parallel mechanism passive planer was designed to accommodate manipulator of lower accuracy. The planer was designed and fabricated for the purpose of optical fiber insertion, which was arbitrarily chosen. Experiment results were shown for the purpose of comparison with the computer simulated model. Kinematics modeling and parameters relationships are to be established further in future works.

Exoskeletons for human performance augmentation are controlled and wearable devices and machines that can increase the speed, strength, and endurance of the operator. So far most researchers focus on the upper limb exoskeletons. To help those who need to travel long distances by feet with heavy loads such as infantry soldiers, this paper presents a control principle of a lower extremity exoskeleton. An exoskeleton foot is designed to measure the human and the exoskeleton's ZMP. Using the measured human ZMP as the reference, the exoskeleton's ZMP is modified by torso control and ground reaction force control so that the exoskeleton can walk stably. Test prototypes, initial experiment results, and preliminary dynamic analysis are also presented.

An investigation was initiated on the study of the kinematic mapping between human hand and the non-anthropomorphic robot hand. To account for the significant dissimilarities between them, a "virtual circle mapping" method has been developed based on the fingertips' relative positions. The principle of the method is to find suitable parameters in the hand frame, transform them to the robot frame, and then compute the robot fingertip positions according to the transformed parameters. The preliminary results and the comparison with the point-to-point mapping results are presented and analyzed. The generalization of the relative-position based mapping idea is also discussed.

Exoskeletons for human performance augmentation are controlled and wearable devices or machines that can increase the speed, strength, and endurance of the operator. So far most researchers focus on the upper limb exoskeletons. To help those who need to travel long distances by feet with heavy loads such as infantry soldiers, this paper presents a control principle of a lower extremity exoskeleton. An exoskeleton foot is designed to measure the human and the exoskeleton's ZMP. Using the measured human ZMP as the reference together with the human leg position signals, the exoskeleton's ZMP is modified by trunk compensation. Test prototypes and initial experiment results are also demonstrated.

This paper focuses on the kinematic modelling, mobility analysis and mechatronics design of a singularityless omni-directional wheeled mobile robot (OWMR). To achieve a singularityless motion, firstly, the kinematic model of a wheeled mobile robot (WMR), a platform equipped with three omni-directional wheels (ODW), is formulated. Based on this model, a smooth trajectory and control command for the 3-wheel OWMR can be obtained. The analysis on the mobility of a WMR is carried out using the functional matrix. It is shown that a WMR with three ODWs has a mobility of three. A WMR with three ODWs was designed and prototyped. Experimental results have been employed to verify the theoretical conclusions.

This paper presents a mechatronic design of a leg mechanism with a flexible gear system. The foot contact with the ground or obstacle is also studied. Simulation and experimental results of the impact effects to the foot and motor are given to illustrate the improved model with the design of a gear-spring motor unit. The impact force is largely low-pass filtered thus protecting the gear box and motor. In this work, a leg mechanism with flexible gear system was designed and constructed. The fuzzy position/force control was also applied. It was found that the fuzzy position/torque control can eliminate the overshoot and works better under different contact velocities. It is also adaptive to unknown obstacles.

This paper focuses on the kinematic modelling, mobility analysis and design of an omni-directional wheeled mobile robots (OWMRs). The composite kinematic models of a wheeled mobile robot (WMR)-a WMR is a collection of the platform and the wheel sub-systems-with a platform equipped with three omni-directional wheels is formulated. The analysis on the mobility of a WMR is carried out using the functional matrix. It is shown that a WMR with three omni-directional wheels has a mobility of three. A prototype of a WMR with three omni-directional wheels are designed, built and tested successfully.

This paper focuses on the gait analysis, design and control of a quadruped legged robot. The structure of the legged robot is first analysed in order to achieve greater mobility and prevent imposing too many constraints for the gait pattern design. Based on all the assumptions made in this study, analysis of the gait pattern is carried out to obtain an optimal regular gait for a legged robot. The kinematic equation of the legs are first formulated and incorporated into the gait simulation program in order to perform the gait simulation. The result of the gait simulation is then input into the control interface program to control the motion of the robot. In the control aspect of this paper, it focuses on the actuator servo control. A prototype of the designed quadruped-legged robot is fabricated and tested successfully based on the above analysis. It is able to move in straight-line motion at around 7 cm per cycle and the cycle took 1 minutes.

The path planning of legged locomotion is difficult in that not only the body trajectory but also the foot placement need to be considered. A general approach of path planning will fail in generating a feasible path for walking machines when facing the huge searching space of legged locomotion. In this paper, an effective method of path planning for 3D walking is introduced. The basic idea is that a feasible path is generated based on the terrain evaluation which produces an index of terrain complexity with respect to the machine mobility. The terrain evaluation is first carried out in a 2D space by considering the leg placement. The 2D terrain complexity is then modified according to the height's influence on the body motion for the 3D results. As an illustration, a feasible path is generated by applying the potential-guided searching technology to the 3D terrain complexity.

As position control alone is not sufficient for machine walking, force control is often suggested. The foot force control allows better force distribution among the legs and therefore reduces the chance of overloading the leg actuators. In this paper, a flexible gear system is designed between the motor and load, by which the force on the gear can be measured and the fuzzy position/force control can then be implemented. The flexible gear system also acts as a buffer to greatly reduce the contact impact magnitude and frequency, therefore the leg can interact with environment with a high speed. Equipped with the flexible gear system in a knee joint, a robot leg has been constructed and an intelligent walking strategy is applied for the leg to walk smoothly on an unknown terrain.

The path generation for legged locomotion systems is difficult in that the legged motion must be considered together with the body movement. Owing to the large searching space resulted from the combination of leg movement and terrain condition, it is quite time consuming to produce an optimal result of body trajectory planning. In this paper, an effective method of body trajectory planning is introduced by virtue of a terrain evaluation that links the terrain condition with the machine mobility. Based on the evaluation, a potential field is constructed for the graph searching. The best first planning (BFP) is adopted to search the optimal path. The path generated with the proposed method could offer the best opportunity to place the machine feet moving with a certain gait over a rough terrain.

When walking on uneven terrain, a walking machine needs to use free gait to deal with the irregularity of terrains. The gait of animals can greatly help the development of free gaits. However, there are few description on the gaits of animal walking on uneven terrain. In the paper, a kinematograph experiment on the moving pattern, of an insect, the cockroach, was introduced. The experiment investigated the behaviour of cockroach walking over different terrain: single obstacle and stairs using high speed cameras. The moving patterns of the cockroach on these terrains are presented and compared in terms of trajectories, displacements, and stick diagrams.

An essential consideration in the development of any motion planning method for a multilegged vehicle is to maintain stability during walking. We propose a leg-end supporting moment (LSM) method, to measure the stability margin of walking robots. The LSM is defined as the product of the force acting on a supporting leg and the distance between the supporting leg and the opposite edge of the support polygon. These forces and distances can be obtained online from the internal robot sensors. The stability margin evaluated by the LSM method, expressed as LSM/sub m/, is defined as the quotient of the leg-end supporting moment over the weight of the robot. If the LSM/sub m/ is greater than zero, the robot is in stable state. Otherwise the robot will be in marginally stable state or will lose its stability. The simulation results show that the LSM method is suitable for the stability analysis in cases with and without external forces acting on the body. The method is not only effective for flat terrain, but also for uneven terrain. These characteristics enable the LSM method to measure the stability margin of a walking robot in the case of drilling, dragging and manipulating, also in the case when the robot is used in forestry.

A method of free gait generation for quadrupeds is presented by introducing the primary/secondary gait. The primary gait is a fixed sequence of leg transfers with modified leg-ends kinematic limits according to the obstacle presence, while the secondary gait is a flexible gait which is generated to adjust the leg-end position. Using the proposed method, a machine tends to move with primary gait during the motion. When the primary gait cannot move the vehicle, the secondary gait is adopted to serve as a complement of the primary gait. Four constraints, namely, stability constraint, kinematic constraint, sequential constraint and neighboring constraint are considered when gaits are generated. The influence of the obstacle is taken into account to modify the gait parameters. With the proposed method based on primary/secondary gait, it is expected to improve the efficiency of free gait generation while maintaining the mobility of the vehicle. Simulation results are given to demonstrate the efficiency of the proposed algorithm.