| Charaterized by high intelligence,efficiency,reliability,and the like,automated guided vehicles(AGV)are widely used in areas of warehouse storage,flexible workshops of industry production,etc.They are used to complete material handling,movement of platforms,and other tasks.As their usage expands continously,the navigation approach of AGV becomes a research hotspot in smart unmaned systems.In industrial production and daily life,AGV needs to adopt interior and outdoor nonstructural dynamic environment,and the traditional navigation approach with fixed guide rail and artificial identification mark is no longer adaptable for such kind of environment independently.Combined with multiple navigation approaches,this paper proposes an AGV navigation approach based on laser radar and visual imaging so as to improve the navigation capability of AGV in non-structural dynamic environment.To achieve the above mentioned research goal,this paper performs detailed researches on the two major steps,namely,the algorithm of simultaneous localization and mapping(SLAM)(perceiving)and the algorithm of path planning(decision making).The specific research contents are as follows:Study and construct the sensor model and measurement error model needed by the AGV navigation approach based on lasor radar and visual imaging,so as to provide corresponding research foundation for SLAM algorithm.Calibrate definite parameters in the measurement error model,determine internal and external parameters among sensors used in the navigation approach,thus guaranteeing the accuracy of SLAM algorithm in the actural navigation test platform.The current single sensor SLAM algorithm is susceptible to interior and outdoor environment change.Aiming at this,this paper proposes a SLAM algorithm with laser and vision tightly coupled: taking the front end speedometer of tightly coupled laser and inertia as the main body,taking the result acquired by the speedometer of tightly coupled laser and vision as the optimizing initial value of attitude estimation.Assign an initial value to the iteration optimizing process of Levenberg Marquardt(LM)of the speedometer of tightly coupled laser and inertia;take error goal function constructed by the three items of speedometer of vision-laser,speedometer of laser-inertia and loop closure detection as the cost item of image optimizing;then a more accurate attitude estimation is obtained via iteration and solution with LM approach.Contrast and analysis of data of the centralizedly performed experiments verify that the integrated algorithm proposed in this paper is better in the consistency of global attitude estimation and the accuracy of local attitude estimation,compared with the single sensor SLAM algorithm that relies only on vision or lasor.The commonly used global path planning A* algrithm is low in planning efficiency,and large in consumption of memory resources,and low in smoothness of path result;besides,the planned path is too close to obstacles.Targeted at these problems and based on JPS algorithm that greatly improves the efficiency of A* algorithm,this paper proposes an improved JPS algorithm that integrates a level fucntion of safe potential field and an improved JPS algorithm with optimized Floyd algorithm.By building a safety level function,a safety level map is constructed via deassigning of the state of grids in the grid map;meanwhile,the introduction of two bias functions of goal and cardinal directions,combined with the safety level function,further reduces time consumption brought by symmetric search,and improves the safety level of the planned path.An second smoothness algorithm process that combines Floyd algorithm is added to obtain an optimal global path node configuration.Compared with A* algorithm,JPS algorithm and other corresponding improved algorithms,contrasts and analysis of simulation test verify that the improved algorithm proposed in this paper is much better in improving planning efficiency,memory resource consumption,smoothness of path result,and path safety,and the like.In complex and dynamic environment,the commonly used dynamic window approach(DWA)is apt to be trapped in local minimum,difficult to obtain planning path with global optimum,and liable to be trapped in second local minimum.Aiming at these problems,the above mentioned improved global path planning algorithm is used to improve DWA algorithm.The obtained path nodes with global optimum are taken as dynamically refreshed goals for DWA,and judging criteria of the second local minimum and replanning framework are set,which guarantee the completeness of path planning and improve the global optimum of the planned path.Contrasts and analysis of simulation test verify that the proposed algorithm can effectively solve the problems of local minimum and the second local minimum caused by dynamic obstacles;meanwhile,it also improves the efficiency of path planning and the capability to seek an optimal path.At last,a test platform of AGV navigation based on laser rader and visual imaging is built to test the actual performance of SLAM algorithm and path planning algorithm in different situations in the test platform.Carry out tests to evaluate the performance of SLAM algorithm in three actual opration environments: large scale scene at school,underground parking garage,and interior long corridor.The test result shows: compared with the laser SLAM algorithm that is commonly used in AGV navigation,the laser and vision SLAM algorithm proposed in this paper is significantly upgraded in the global consistency of attitude estimation and the accuracy of mapping.The effectiveness of the improved path planning algorithm is verified in the safety level grid map projected by the point cloud map of interior long corridor via SLAM algrithm.The test result shows that the efficiency and quality of path planning improves significantly.Synthesizing the test results of SLAM algorithm and path planning algorithm in actual AGV navigation platform,it proves that the AGV navigation approach based on laser radar and visional imaging in this paper is feasible and reliable in multiple different interior and outdoor scenes. |
