Autonomous Navigation Using Deep Reinforcement Learning in ROS

For an autonomous robot to move safely in an environment where people are around and moving dynamically without knowing their goal position, it is required to set navigation rules and human behaviors. This problem is challenging with the highly stochastic behavior of people. Previous methods believe to provide features of human behavior, but these features vary from person to person. The method focuses on setting social norms that are telling the robot what not to do. With deep reinforcement learning, it has become possible to set a time-efficient navigation scheme that regulates social norms. The solution enables mobile robot full autonomy along with collision avoidance in people rich environment.

Multi-Satellite Relative Navigation Scheme for Microsatellites Using Inter-Satellite Radio Frequency Measurements

The inter-satellite relative navigation method—based on radio frequency (RF) range and angle measurements—offers good autonomy and high precision, and has been successfully applied to two-satellite formation missions. However, two main challenges occur when this method is applied to multi-microsatellite formations: (i) the implementation difficulty of the inter-satellite RF angle measurement increases significantly as the number of satellites increases; and (ii) there is no high-precision, scalable RF measurement scheme or corresponding multi-satellite relative navigation algorithm that supports multi-satellite formations. Thus, a novel multi-satellite relative navigation scheme based on inter-satellite RF range and angle measurements is proposed. The measurement layer requires only a small number of chief satellites, and a novel distributed multi-satellite range measurement scheme is adopted to meet the scalability requirement. An inter-satellite relative navigation algorithm for multi-satellite formations is also proposed. This algorithm achieves high-precision relative navigation by fusing the algorithm and measurement layers. Simulation results show that the proposed scheme requires only three chief satellites to perform inter-satellite angle measurements. Moreover, with the typical inter-satellite measurement accuracy and an inter-satellite distance of around 1 km, the proposed scheme achieves a multi-satellite relative navigation accuracy of ~30 cm, which is about the same as the relative navigation accuracy of two-satellite formations. Furthermore, decreasing the number of chief satellites only slightly degrades accuracy, thereby significantly reducing the implementation difficulty of multi-satellite RF angle measurements.

Recovery and guidance technology of collision network based on multi-source information fusion

To solve the problem of low precision of pose estimation and poor anti-jamming in the recovery process of small Shipborne UAV, a multi-sensor navigation scheme was proposed, based on vision / IMU / GPS fusion. Firstly, an integrated navigation scheme of multi-sensor fusion in collision recovery guidance was proposed. GPS is used for long-distance guidance. Vision / IMU / GPS fusion guidance is used in approach phase. GPS and visual information are fused based on extended Kalman filter to obtaion position information. IMU and visual information are fused based on least square method to obtain the optimal attitude information. According to this method, the optimal pose information is obtained to improve the guidance accuracy and the success rate of net collision recovery.The simulation results show that the guidance deviation is reduced by 30% compared with the unfused data.

MODEL OF WORKING SHIP CROSSING CHANNEL

An application method for working ship crossing safely is proposed to determine how to make navigation scheme at a certain time. This method makes it possible for decision makers to make reasonable judgments at different times. In this paper, the position relationship between working ship and navigation vessel in waterway is analysed by considering the ship size, hydrological conditions of waterway, ship arrival model and ship navigation trajectory. Using genetic algorithm, the operation scheme of keeping a safe distance between the working ship and the vessel in the channel is solved by taking the speed and direction of the working ship as genetic factors. By analysing the crossing scheme at each starting time in a given time range, the optimal crossing scheme with the farthest distance between the working ship and the vessels in the channel is obtained. According to the measured data, the simulation is carried out with MATLAB to verify the model of working ship crossing channel. The results show that it is safe and reliable to choose the navigation scheme proposed in this paper, which has strong application value.

Transverse SINS/DVL Integrated Polar Navigation Algorithm Based on Virtual Sphere Model

Due to the principle error caused by the Earth’s sphere model in transverse polar navigation, the ellipsoid model is usually used for transverse transformation. In order to avoid the complex transverse transformation of the ellipsoid model, a virtual sphere model, which can simplify the transformation of the ellipsoid model, is constructed in this paper. With the requirements of high accuracy and long-time navigation for AUV in the polar region, the integrated navigation scheme of SINS/DVL, which is based on the virtual sphere model, is proposed. The proposed method can improve the navigation accuracy and suppress the oscillation error. The error equations of transverse SINS are established based on the virtual sphere model. Then, the transverse SINS/DVL integrated navigation algorithm is derived according to the new error equations. The simulation results show that the navigation accuracy of the proposed method is equivalent with the traditional ellipsoid model method and is better than that of the traditional sphere model method. However, the complexity of the proposed method is simpler than the traditional ellipsoid model method. Moreover, it is verified that the navigation accuracy of SINS/DVL integrated navigation system based on the virtual sphere model meets the requirements of AUV.

Landmark-Based Inertial Navigation System for Autonomous Navigation of Missile Platform

As a new information provider of autonomous navigation, the on-orbit landmark observation offers a new means to improve the accuracy of autonomous positioning and attitude determination. A novel autonomous navigation method based on the landmark observation and the inertial system is designed to achieve the high-accuracy estimation of the missile platform state. In the proposed method, the navigation scheme is constructed first. The implicit observation equation about the deviation of the inertial system output is derived and the Kalman filter is applied to estimate the missile platform state. Moreover, the physical observability of the landmark and the mathematical observability of the navigation system are analyzed. Finally, advantages of the proposed autonomous navigation method are demonstrated through simulations compared with the traditional celestial-inertial navigation system and the deeply integrated celestial-inertial navigation system.