As there come to be more applications of intelligent robots, their task object is becoming more varied. However, it is still a challenge for a robot to handle unfamiliar objects. We review the recent work on the feature sensing and robotic grasping of objects with uncertain information. In particular, we focus on how the robot perceives the features of an object, so as to reduce the uncertainty of objects, and how the robot completes object grasping through the learning-based approach when the traditional approach fails. The uncertain information is classified into geometric information and physical information. Based on the type of uncertain information, the object is further classified into three categories, which are geometric-uncertain objects, physical-uncertain objects, and unknown objects. Furthermore, the approaches to the feature sensing and robotic grasping of these objects are presented based on the varied characteristics of each type of object. Finally, we summarize the reviewed approaches for uncertain objects and provide some interesting issues to be more investigated in the future. It is found that the object’s features, such as material and compactness, are difficult to be sensed, and the object grasping approach based on learning networks plays a more important role when the unknown degree of the task object increases.

This paper investigates the dynamic parameter identification for flexible joints without additional sensors. The estimation accuracy of traditional methods based on the original dynamic model is limited by measurement of angular acceleration. Recently, approaches using integral operations are raised to solve this problem. However, the integral operations easily result in error accumulation and deteriorate the parameter estimation accuracy. This study proposes a new dynamic parameter identification method for flexible joints. By constructing an alternative system function and estimating its outputs through non-model-based adaptive control, we avoid the measurement of angular acceleration and integral operations. On this basis, dynamic parameters of flexible joints are estimated using the recursive least squares (RLS) method. The experimental results indicate that the proposed method possesses better estimation accuracy than the general RLS- and integral-based methods. It is also validated that the proposed method can provide accurate estimation even under a relatively low-resolution system. Applications to model-based motion control and physical human-robot interaction are also verified.

This paper proposes a hexapod robot posture control method for rugged terrain to solve the problem of difficulty in realizing the posture control of a foot robot in rough terrain. The walking gait and original position of a six-legged robot is planned, and the Layer Identification of Tracking (LIT) strategy is developed to enable the robot to distinguish mild rugged terrain and severe rugged terrains automatically. The virtual suspension dynamic model is established. In mild rugged terrain, the posture maintenance strategy is adopted to keep the stability of the torso. In severe rugged terrain, the posture adjustment strategy is adopted to ensure the leg workspace and make it more widely adapt to the changing terrain, and a gravity center position adjustment method based on foot force distribution is designed to use foot force as feedback to control the position and attitude. The experiment of posture control in rough terrain and climbing experiment in the ladder terrain shows that the hexapod robot has good posture maintenance and posture adjustment effects when traversing complex terrain through the posture maintenance strategy and the posture adjustment strategy. Combined with the terrain identification method based on LIT, the hexapod robot can successfully climb the ladder terrain through the identification of the changing ladder terrain, and the movement of the posture adjustment process is stable.

In manufacturing, traditional task pre-programming methods limit the efficiency of human–robot skill transfer. This paper proposes a novel task-learning strategy, enabling robots to learn skills from human demonstrations flexibly and generalize skills under new task situations. Specifically, we establish a markerless vision capture system to acquire continuous human hand movements and develop a threshold-based heuristic segmentation algorithm to segment the complete movements into different movement primitives (MPs) which encode human hand movements with task-oriented models. For movement primitive learning, we adopt a Gaussian mixture model and Gaussian mixture regression (GMM-GMR) to extract the optimal trajectory encapsulating sufficient human features and utilize dynamical movement primitives (DMPs) to learn for trajectory generalization. In addition, we propose an improved visuo-spatial skill learning (VSL) algorithm to learn goal configurations concerning spatial relationships between task-relevant objects. Only one multioperation demonstration is required for learning, and robots can generalize goal configurations under new task situations following the task execution order from demonstration. A series of peg-in-hole experiments demonstrate that the proposed task-learning strategy can obtain exact pick-and-place points and generate smooth human-like trajectories, verifying the effectiveness of the proposed strategy.

This paper focuses on motion planning for mobile manipulators, which includes planning for both the mobile base and the manipulator. A hierarchical motion planner is proposed that allows the manipulator to change its configuration autonomously in real time as needed. The planner has two levels: global planning for the mobile base in two dimensions and local planning for both the mobile base and the manipulator in three dimensions. The planner first generates a path for the mobile base using an optimized A* algorithm. As the mobile base moves along the path with the manipulator configuration unchanged, potential collisions between the manipulator and the environment are checked using the environment data obtained from the on-board sensors. If the current manipulator configuration is in a potential collision, a new manipulator configuration is searched. A sampling-based heuristic algorithm is used to effectively find a collision-free configuration for the manipulator. The experimental results in simulation environments proved that our heuristic sampling-based algorithm outperforms the conservative random sampling-based method in terms of computation time, percentage of successful attempts, and the quality of the generated configuration. Compared with traditional methods, our motion planning method could deal with 3D obstacles, avoid large memory requirements, and does not require a long time to generate a global plan.