Research On Dynamic Path Planning Methods For Dual Manipulators System In Coordinative Mode

With the increasing complexity of industrial tasks and the need for intelligence,dual manipulators systems are getting more and more widely used in automated production lines.Under coordinative working mode,a dual manipulators system needs to plan dynamic collision-free paths for all the manipulators in order to avoid collisions during movement.There still exist numerous limitations in traditional path planning algorithms: due to the dynamic construction of node maps,the map search method proved to be inefficient;the sampling directions of the Rapidly-exploring Random Tree algorithms and the sampling points of the Probabilistic Roadmap Methods are both have large randomness,yielding paths that are lacking in optimality and stability.What’s more,the path curves generated by the above algorithms are all poor in flexibility,and the traditional path planning algorithms cannot achieve the initial path regression after obstacle avoidance when the system already has initial paths.This thesis proposes solutions to address each of these limitations with the following main research topics.:(1)Aiming at the problem of inability to achieve real-time dynamic collision prediction of dual manipulators,this thesis proposes a real-time prediction method for the minimum distance between dual manipulators.Based on the Sphere Sweep Volume spatial geometry collision detection algorithm,the spatial coordinates and the Jacobi matrixes of arbitrary point on the central axis of the manipulator are firstly derived,and then the fitting prediction formula of the minimum distance between the dual manipulators is derived using the differential motion analysis of the spatial geometry of the manipulator arm rod.The formula can achieve collision prediction solely through vector operations,which with high computational efficiency.Cases study show that the prediction error of the minimum distance variation value is within 5% relative to the sampling step length,except for the singular relative position points,with high fitting accuracy.Tests on the execution efficiency of the prediction method also show that all execution times are within 1ms,meeting the realtime requirements of industrial applications.(2)Aiming at the incompatibility of traditional algorithms in terms of efficiency and stability,this thesis proposes a dynamic path planning method for dual manipulators based on the sampling coordinate system.This method discretizes the entire path planning cycle,determines the alternative sampling point sets for individual manipulator arm by constructing a sampling coordinate system within a single sampling cycle,then combines the alternative sampling point sets of different manipulators into a complete local sampling node map,and uses the penalty function of the A* algorithm to select the optimal path point as the planning starting point for the next planning cycle.Numerous case studies have demonstrated that this method reduces the total number of path offsets by 20% to 40% and the execution time of the algorithm by 90% compared to traditional path planning algorithms.(3)Aiming at solving the problem of low flexibility of the planned path curves and the inability to achieve optimal planning in the local sampling domain,this thesis proposes a dynamic path smoothing method based on a mathematical model of the dual manipulators system.This method constructs a mathematical model of a dual manipulators flexible path planning based on three assessment parameters,namely path reachability,path length and path flexibility,and solves the mathematical model by simplifying it with the real-time prediction formula.Based on a large number of case studies,the mathematical model based dynamic path smoothing method can improve the path smoothing by over 70% compared to the sampling-based dynamic path planning method,and also significantly decrease the curve length and improve execution efficiency.(4)To realise the dynamic path regression of a dual manipulators system with initial task paths after avoiding obstacles,an energy conversion based dynamic regression method is proposed.This method establishes an energy body model for individual manipulator and a collision virtual elastic potential energy model between the two manipulator arms,and then transfers the virtual energy to each energy body model according to a certain distribution criterion based on the law of energy conservation.The collision virtual energy is transformed into the corresponding initial path curve deformation using the elastic deformation potential energy calculation formula,thus achieving a perfect docking between collision avoidance and path regression.This method takes an energy perspective on path planning based on the path curve as a whole.The principle is simple,the algorithm is executed efficiently,and the regression of the path curve is achieved while maintaining the flexibility and optimality of the planned path.(5)The test platform is built in the background of material handling in the coordinated working mode of the dual manipulators system,and its composition and principles are introduced in three aspects: hardware framework,software composition and communication control technology.Finally,based on the test platform,a real-life replication experiment of the algorithm simulation cases and a dual manipulator coordinated material handling experiment is performed.The real-world replication experiment verifies the effectiveness of the proposed methods in this thesis,and the dual manipulators coordinative material handling experiment demonstrates the adaptability of the proposed methods to different random working conditions.