| Landing on celestial bodies surface and exploring is a effective research method in deep space exploration. Due to the signal delay produced by distance between the earth and deep space body, and the short duration of descent, the spacecraft must equip autonomous guidance navigation and control system. And it is a critical technology for deep space exploration mission. With the supports of the 863 Program'Autonomy Technology of Deep Space Exploration and its Simulation and Demonstration System'and National Natural Science Foundation of China'Theory and Method of Deep Space Autonomous Navigation', this dissertation deeply studies the autonomy navigation theoretics and key methods of deep space exploration landing phase. The main contents of this dissertation are as follows:Firstly, according to the characteristic of landing phase the feature points tracking algorithm is studied and its performance is tested by carrying out simulation. Then the crater recognition method based on the edge is proposed which is the groundwork for autonomous navigation since the crater is the distinct landmark on surface. This method employs the modification edge detecter to detect the potential crater edge then tensor voting algorithm to select the salience points to reduce the cost of compute. Curve fitting acquires the position and semimajor axis estimation of crater from RANSAC (random sample consensus) algorithm.Secondly, the vision based autonomous navigation algorithm is studied, which adopts the batch method based on corresponding feature points. The epipolar geometry based motion estimation is deeply analyzed for deep space mission. Then the probe relative motion estimation and landing sites assessment algorithm is presented. Due to the planar surface of landing site, the motion estimation algorithm based on essential matrix is instability. Motion Estimation algorithm from Homography is employed to fix this problem. Homography matrix is calculated from tracking at least 4 feature points through images by least squares method. To estimate the motion of probe, the distance between probe and surface is also obtained by laser altimeter. The Landing Site is detected by the mean square error of the image intensity. Adding the motion of probe the algorithm determines the gradient of the site.Next, this thesis proposes a combine algorithm for use by a deep space exploration spacecraft to estimate the absolute position and attitude, relative to celestial body fixed frame, on broad during the descent phase. This algorithm is composed of the relative motion recovery which provides part motion states estimates based on tracing feature through the monocular image sequence, and landmark (crater) matching algorithm which supplies the distance between two craters or parameters of crater to find the scale of the relative motion and absolute position of spacecraft. This algorithm enable motion estimate when the spacecraft descents and observes less craters.Then the guidance and control strategy is studied, which enable the probe achieving pinpoint planetary landing. A landing guidance law based on predictive fuzzy control is studied. This method guides the probe descent according to the predictive final state. Since the target celestial body surface is unknown until landing, the probe must face the uncertainty gravitation in deep space mission. Hence the adaptive and robust control algorithm is designed for solving this problem, which can effectively estimate the uncertainty brought by error of navigation system.Lastly, combining the 863 Program'Autonomy Technology of Deep Space Exploration and its Simulation and Demonstration System'and National Natural Science Foundation of China'Theory and Method of Deep Space Autonomous Navigation', semi-physical simulation system is built up to simulate the soft landing autonomous navigation algorithm, which is based on optical navigation camera and 3-D synthetic terrain. The feasibility of several navigation algorithm developed above are confirmed by the system.
|
PDF Full Text Request