Visual SLAM (VSLAM) has been developing rapidly due to its advantages of low-cost sensors, the easy fusion of other sensors, and richer environmental information. Traditional visionbased SLAM research has made many achievements, but it may fail to achieve wished results in challenging environments. Deep learning has promoted the development of computer vision, and the combination of deep learning and SLAM has attracted more and more attention. Semantic information, as high-level environmental information, can enable robots to better understand the surrounding environment. This paper introduces the development of VSLAM technology from two aspects: traditional VSLAM and semantic VSLAM combined with deep learning. For traditional VSLAM, we summarize the advantages and disadvantages of indirect and direct methods in detail and give some classical VSLAM open-source algorithms. In addition, we focus on the development of semantic VSLAM based on deep learning. Starting with typical neural networks CNN and RNN, we summarize the improvement of neural networks for the VSLAM system in detail. Later, we focus on the help of target detection and semantic segmentation for VSLAM semantic information introduction. We believe that the development of the future intelligent era cannot be without the help of semantic technology. Introducing deep learning into the VSLAM system to provide semantic information can help robots better perceive the surrounding environment and provide people with higher-level help.

Continuum robots with redundant degrees of freedom and postactuated devices are suitable for application in aerospace, nuclear facilities, and other narrow and multiobstacle special environments. The development of a snake-inspired continuum robot is presented in this study. The morphological skeleton structure of the snake body is simulated using underactuated continuum joints, which include several rigid-body joints in series. Each rigid-body joint is driven by the traction of a wire rope. Based on the layered-drive principle, angular synchronous motion can be realized in space with multiple rigid-body joints in a single continuous joint, which can considerably reduce the complexity of the inverse kinematics solution, terminal drive box, and control system. The static and dynamic characteristics of the snake-inspired robot are obtained through torque balance and an equivalent transformation. Finally, we demonstrate trajectory planning and load capacity testing in two robot prototypes with arm lengths of 1500 and 2300 mm (including two and four continuous joints, respectively). The rationality of the structure and the correctness of the control of the layered-drive snake-inspired robot are verified.

Insects that can perform flapping-wing flight, climb on a wall, and switch smoothly between the 2 locomotion regimes provide us with excellent biomimetic models. However, very few biomimetic robots can perform complex locomotion tasks that combine the 2 abilities of climbing and flying. Here, we describe an aerial–wall amphibious robot that is self-contained for flying and climbing, and that can seamlessly move between the air and wall. It adopts a flapping/rotor hybrid power layout, which realizes not only efficient and controllable flight in the air but also attachment to, and climbing on, the vertical wall through a synergistic combination of the aerodynamic negative pressure adsorption of the rotor power and a climbing mechanism with bionic adhesion performance. On the basis of the attachment mechanism of insect foot pads, the prepared biomimetic adhesive materials of the robot can be applied to various types of wall surfaces to achieve stable climbing. The longitudinal axis layout design of the rotor dynamics and control strategy realize a unique cross-domain movement during the flying–climbing transition, which has important implications in understanding the takeoff and landing of insects. Moreover, it enables the robot to cross the air–wall boundary in 0.4 s (landing), and cross the wall–air boundary in 0.7 s (taking off). The aerial–wall amphibious robot expands the working space of traditional flying and climbing robots, which can pave the way for future robots that can perform autonomous visual monitoring, human search and rescue, and tracking tasks in complex air–wall environments.

The majority of sprawling-posture quadrupedal vertebrates, such as geckos and lizards, adopt a cyclical lateral swing pattern of their trunk that is coordinated with limb movements to provide extraordinary flexibility and mobility. Inspired by the gecko's locomotory gait and posture, a gecko-like robot with a flexible spine driven by shape memory alloy (SMA) springs was proposed in this work. The static parameters of the SMA spring were experimentally measured, and the flexible spine driven by SMA springs can be deflected bidirectionally. A kinematic model of the spine mechanism was established, and the mathematical relationship between the thermodynamic behavior of the SMA springs and spinal deflection was systematically analyzed. When a gecko trots with a lateral swing pattern of its trunk, the body and the spine show a standing wave shape and a single-peak C-type curve, respectively. The lateral spine deflection and trotting gait were combined in a collaborative model of a flexible spine and limbs to describe in detail the specific relationships between leg joint variables and spine deflection angle. Planar motion tests of a prototype robot were also conducted by using four high-speed cameras to record the trajectory of each point of the body, which verified the proposed model. From the acquired results, it was demonstrated that, compared with a rigid body, a robot with a flexible spine has a longer stride length, higher speed, and a greatly reduced turning radius.