Simultaneous Localisation And Mapping For Mars Rover

With the development of space technology and universal exploration,the autonomous navigation technology is challenged by the space missions.The autonomous navigation technology of rover is a key component of planetary exploration missions,which also has a direct influence on the reliability and scientific return.The traditional navigational method by rover based on inertial navigation no longer satisfies the needs of furure plan-etary exploration missions.SLAM(Simultaneous Localization and Mapping)algorithm is a new type of navigation system in planetary surface exploration missions,which is a significant support for planetary surface navigation.Therefore,it is necessary to develop a new generation of autonomous navigation method for Mars rover based on SLAM tech-nology.With the support of the National Natural Science Foundation of China,'Coopera-tive Navigation and Control for Long-Range Rover Exploration of Mars Surface',and the project of Aerospace System Engineering Institute of Shanghai,'Development of proto-type of Mars rover for autonomous navigation ground test',the thesis is focused on the SLAM algorithm of Mars rover.The content of the thesis can be expressed into the fol-lowing five aspects:Firstly,the computational complexity of SLAM algorithm of Mars rover based on EKF framework is studied.The computational complexity will increase rapidly with the development of exploration process and the increase of map area.In the thesis,the ex-pression structure and influencing factors of computational complexity are derived.The computational complexity of the EKF-SLAM is proportional to the square of the number of feature points contained in the state variables in a single filtering process.Likewise,the computational complexity of the overall algorithm is proportional to the cube of the total number of feature points contained in the state variables.Considering this problem,a SLAM algorithm based on the local submaps of Mars rover is presented.An indepen-dent local submap is established for the observed feature points.When the number of the feature points contained in the local submap reaches a certain threshold value,the local submap is integrated into the global map.Compared to traditional EKF-SLAM algorithm,the accuracy and complexity of our proposed algorithm are better.Secondly,the consistency of the SLAM algorithm of Mars rover based on EKF frame-work is studied.The recursive solution of navigation problem and the estimation process of uncertainty of rover and land marks are given by EKF-SLAM algorithm directly.How-ever,when solving practical problems,the linearization of the system observation model will lead to non-convergence for nonlinear system.The linearization error cannot guar-antee the matching of the EKF estimation covariance with the real covariance.Therefore,the problem of consistency of the estimated results of the algorithm is generated.The tra-ditional EKF-SLAM algorithm is established in the absolute coordinates,which increases the linearization error,due to the accumulative uncertainty of the rover's pose and the ac-cumulative uncertainty of the feature point's position are introduced into the observation equation.Therefore,a body-fixed Mars rover SLAM algorithm is presented in the thesis.The feature points are updated in the coordinates of Mars rover.Uncertainty of the rover's attitude and the feature points' position will not be introduced into the linearization of the observation equation.The uncertainty of observation process can be reduced by reducing the uncertainty of residual.The variation of rover's pose is added to the state quantity as an independent feature point,and then the EKF update is carried out.Finally,the posi-tion of feature points in the coordinates of rover is updated.The simulation results also demonstrated the effectiveness of proposed algorithms.Thirdly,the searching algorithm of overlapping area between multiple Mars rovers is studied.The map merging problem is the key component in realizing the long-range exploration,meanwhile the searching algorithm of the overlapping area is a pre-step.Con-sidering the sparsity of image descriptior,an searching algorithm of overlap area based on improved Bag-of-Word model is presented.The descriptor is generated according to the dictionary,which is used to search similar parts in different sets of sequence images.The similarity scoring function based on L1 norm form is established.L1 norm solution is usu-ally sparse and tends to choose some very large values with fewer numbers or some very small values with more numbers,which can represent the degree of difference between two images better.The relative threshold of the similarity scoring function is set,and the similarity inhibition algorithm is used to remove the low similarity image according to the similarity score graph.The simulation results are discussed by a series of simulation experiments.Fourthly,research is focused on map merging and pose merging between multiple rovers.Considering the large amount of calculation of images and the computing power ofonboard computer,a map merging algorithm based on sparse feature points is proposed.A large-range feature point map is constructed by merging several small sparse feature point maps.The sparse feature point map merging problem is transformed into calculating the pose transformation relationship of the overlapping areas of the two Mars rovers,which avoids the direct use of ICP algorithm.Finally,the thesis conducted a complete research and simulation verification SLAM and mapping algorithm of Mars rover.Lastly,a prototype of Mars rover is built to carry out ground tests of the navigation system of the Mars rover.The software and hardware of the navigation system are intro-duced.Meanwhile,the global planning algorithm based on A* is tested,and the results show that the method can improve the path search efficiency greatly.And then the local path planning and obstacle avoidance algorithm based on VFH+ are tested.The results show that the method can effectively carry out local path planning and estimate the planned trajectory with high smoothness.Finally,the SLAM algorithm of Mars rover are tested,and the results show that the method can estimate the pose of Mars rover well.The effec-tiveness and accuracy of the navigation system of Mars rover are proved by the ground tests and the analysis results.