Research On SLAM Algrithm For Robot MaNSoC Dedicated Chip

Robot Simultaneous Localization and Mapping(SLAM)uses specific sensors carried by mobile robots to build a model of the environment during the movement without prior information about the environment,and at the same time estimate its own movement state.Due to power consumption or size constraints,mobile platforms are often equipped with embedded systems and high-performance computers are not allowed.Especially in the field of robot positioning,mapping and navigation,most mobile robots in my country are equipped with general-purpose embedded chips.It has low mapping efficiency,poor real-time positioning,low integration,high power consumption,high cost,and the corresponding algorithms and chips cannot be efficiently adapted.my country's robots are relatively backward in terms of chip technology,software and hardware modules,and lack an organic combination of key technologies such as dedicated chips and core algorithms.The core part of the SLAM algorithm-multi-sensor fusion algorithms often have defects such as high computational complexity and poor robustness.The bottleneck of the dedicated robot chip and the defects of the multi-sensor fusion algorithm severely limit the accuracy and computational efficiency of the SLAM algorithm.Mapping and Navigation System on Chip(MaNSoC)is a heterogeneous multi-core system on chip.It is equipped with a dedicated chip for robot positioning,mapping and navigation.This topic is oriented towards the robot MaNSoC special chip,considering the shortcomings of the current multi-sensor fusion algorithm,starting from the most popular multi-sensor fusion algorithm-extended Kalman filter(EKF),and deeply analyzing the limitations of the traditional EKF SLAM algorithm: complex environment The computational complexity of the algorithm caused by the next large-scale map is increased;the impact in the dynamic environment causes the degradation of the robot sensor's perception,causing inconsistency in the estimation.Aiming at the former,this subject proposes an EKF SLAM software and hardware acceleration algorithm for MaNSoC dedicated chip: application of software and hardware co-design theory,based on detailed computational complexity analysis and rigorous mathematical derivation,optimized from two levels of software and hardware And to meet the design constraints,and verify the effectiveness of the algorithm through a heterogeneous system-on-chip(ARM+FPGA).Aiming at the latter,this topic proposes an alternative solution: an anti-shock scanning matching algorithm for MaNSoC dedicated chips: based on multi-core CPU and a dedicated accelerator for scanning matching algorithms,it improves the two shortcomings of traditional scanning and map matching algorithms: improved The accuracy of a priori estimation and the real-time nature of the scan matching algorithm.This paper verifies the effectiveness of the algorithm through a robot platform equipped with a MaNSoC dedicated chip.The anti-impact scanning matching algorithm for heterogeneous systems-on-a-chip proposed in this subject has been commercialized and is currently used in many companies' robot cleaners,with a total of more than 1million units.