Visual SLAM has been the hot research topic in recent years, which treats visual image as processing objects. Deep learning shows the prominent advantages in image processing, which makes it possible to combine the visual SLAM and deep learning. The characteristics and properties of traditional SLAM and SLAM based on deep learning are summarized. The prominent achievements on visual odometry and loop closure detection incorporated with deep learning are introduced. The future research directions of advanced SLAM based on deep learning are discussed.

In order to improve the accuracy of spatial direction measuring instrument, and to meet the requirement of extremely high measurement accuracy, the influence of image sensor’s pixel displacement on centroiding accuracy is not negligible. The laser interferometry is a major method for measuring pixel displacement in current. The causes and effects of image sensor’s pixel displacement are described in this paper. The techniques of laser interferometry for measuring pixel displacement are introduced and analyzed. The key issues, difficulties and future development trends of the technique for pixel displacement measurement are summarized.

The cluster spacecraft will play an important role in future space missions due to its unique advantages, and its boundary control becomes a hot spot for research. In this paper, based on the cluster spacecraft’s swarm control model，the particle swarm optimization algorithm is used to acquire the boundary parameters(the minimum envelope radius and the spherical coordinates) of the cluster spacecraft when the state are stable. At the same time, the control method of the spherical cavity potential function is proposed. Combined with the feedback information of the boundary parameters of the cluster spacecraft, the spherical boundary control of the cluster spacecraft is realized. The feasibility of the algorithm is verified by simulation.

The operation of space flexible joint space manipulator relates to many factors. One of the key factors is the elastic vibration of the joints. A control strategy is introduced for the trajectory tracking of a twolink space flexible joint manipulator. The EulerLagrange formula is used for modeling the dynamics of the manipulator system. Then, a successive linearization model predictive control scheme is proposed for the tracking control of the space manipulator. Finally, some numerical simulation are carried out to validate the effectiveness of the proposed method.

 Aiming at the characteristics of the guidance, navigation and control (GNC) system of the lunar rendezvous and docking, a safety auxiliary prediction system based on the humanmachine collaboration frame is designed. The system includes two parts: building the rendezvous and docking safety rule knowledge base by the machine learning method, and predicting the safety by the use of the base. Firstly, in the ground semiphysical simulation test environment, the features are extracted and matched, while the object is imaged by the navigation surveillance camera. And then, the matched features and rendezvous and docking bias are used to form the training samples pairs. Secondly, the safety rules knowledge base is used to predict the safety of the subsequent process rendezvous and docking, according to the image of current state, and obtain the probability of success. The semiphysical experiments on the ground show that the system can improve the intelligence level in the flight control process and give effective data and monitoring support for the ground flight controllers.

 The operating system is the basic software necessary for the spacecraft. The reliability and security of the operating system are directly related to the success or failure of the aerospace model mission. Although various methods have been used to guarantee the reliability and security of the operating system, there are still cases where the defects cannot be completely eliminated. Therefore, it is imperative to conduct formal verification research on space operating systems. The verification of requirement layer is a part of the formal verification of the operating system. Based on the analysis of the operating system requirements,a method of finite state machine is used to formally describe in the demand layer and applied to a spacecraft. SpaceOS2 is modeled at the requirements level, and correspondingly described in the theorem proving tool Coq; then formal definitions are made for some global properties, and the machinecheckable proof that the system model satisfies these properties is given. The proof result shows that the finite state machine method is feasible to formally verify the operating system requirements layer and lays the foundation for further comprehensive formal verification.

The simulation of the timeofflight sensor is mainly used to evaluate the processing algorithm of the depth data. Existing simulation models include two types,one is based on a specific noise pattern generated from a threedimensional model of a scene,and the other one is a physical process simulation model based on optical signal propagation. The simulation model of timeofflight imaging is implemented based on physical processes of optical signal propagation. The system is oriented to the needs of engineering applications.The full physical process of timeofflight imaging is modeled, and the simulation parameters are set according to specific sensors.The depth data of the scene is generated the impact of parameter changes on depth data is evaluated and this model is used to analyze typical errors in timeofflight imaging.

Based on an electrostatic forcebalance quartz acceleraometer, the dynamic model is established and a closeloop control system based on electrostatic force is designed. The stiffness model and deflection model of the pendulous reed are established, indicating that the nonuniform of the capacitor gap variable quantity affects the drive and sense parameters. Besides, the electrostatic force model of drive capacitor is established, and the effect of negative stiffness on the dynamic model is analyzed. To decrease the mechanical thermal noise, the accelerometer is evacuated, which causes the system damping decreasing to zero. To avoid oscillation near the system resonant frequency, a damping compesator is added into the closeloop system. Besides, an integal element and a correction element are designed to increase the rapidity and atablility of the whole system. The simulation results show that, the bandwidth is 2 380 Hz, and the overshoot abd response time are 3% and 0.5 ms respectively. Besides, the closeloop system can work stably with the increasing of negative stiffness.