Graphic-Based Interactive Path Planning for Large-Scale Bridge Maintenance Cranes
Graphic-Based Interactive Path Planning for Large-Scale Bridge Maintenance Cranes
Abstract
Crane-type manipulators are an essential rigging system for bridgemaintenance. The control of the bridge maintenance crane is mostly executedrelying on operators' manual skills and experience. This is because automation ofthe manipulator control is difficult especially in an unstructured constructionenvironment. Telerobotic operation allows for implementing a human-machineinterface, and may serve better than full automation for efficient manipulatorcontrol. Off-line path planning provides a guidance that promotes efficientimplementation of the telerobotic operation on the construction site. This paperpresents the architecture of the graphic-based path planner for creating themanipulator path from a starting position to a desired destination in a CADenvironment. The path planning supports a graphic overlay control that enableshuman-machine interface during telerobotic operation.
Introduction
Bridge paint maintenance requires a special rigging system because of theelevated work place. Crane-type manipulators are widely used to provide accessto bridge structures for inspection, paint removal and application, etc. Thesuspended working area is an unsafe working environment. Sometimes the debrisof toxic lead paint on the bridge beam surface causes a health hazard to humanworkers.
When the bridge maintenance process is robotized in an attempt to improvesafety and health conditions, automation of the crane control is desired. Fullautomation is, however, difficult in an unstructured construction environment. Asan alternative approach, a control mode based on human-machine interface can beused to replace manual control. The interface creates a human-in-the-loop controlthat integrates the human operator's decision capability with the machinecontroller's automation capability during the manipulator control. Off-line pathplanning provides a basis for on-line manipulator control in telerobotic operation. Predetermination of manipulator path can generate control data required fordeploying the crane under the bridge.
A graphic-based path planner is developed in a CAD environment as a partof the Robotic Bridge Maintenance project for the North Carolina Department ofTransportation (NC DOT). The planner is designed to facilitate the creation of themanipulator path from a starting position to a desired destination throughoperator-interface. The data from the planning are used for a graphic overlaycontrol that enables efficient human-machine interface. This paper presents thearchitecture of the graphic-based off-line path planning.
Off-line Path Planning for Robotic Manipulator Control
In automatic manipulator control the off-line path provides a means forsimulating manipulator motion without actual operation of the equipment. Thedata from the planning is used for guiding a manipulator during on-linemanipulator control in a real situation (Craig 1986). The computer path plannercreates via points, joint angles, velocities as well as acceleration that describe atrajectory in a permitted work space. Coupled with the real-time data input from avariety of sensors, these data can be directly used during the telerobotic control ofconstruction manipulators.
For automated path planning the operator needs only to specify the desiredgoal position and orientation of the end-effector. The path planner thendetermines the shape of the path to the goal position, manipulator configuration toreach via points, duration, velocity, etc. Fundamental analytical tools for solvingmotion-planning problems have been studied in various approaches such asconfiguration-space framework (Lozano-Perez 1983), free-space framework(Brooks 1983), and a generalized Voronoi diagram (GVD) (Lee et al. 1981).
Off-line path planners built on a graphic simulation capability allowmanipulator motions to be displayed through graphic representation. This is anattempt to link the path planner to CAD data bases. The feature data of CADdrawings which were generated by the designer of a part or subsystem can bedirectly used by the path planner (Craig 1986). The simulation offers anopportunity to visually inspect the results of the path planning and constraints thatmay be encountered during the navigation of the manipulator.
When implementing any automated manipulator control in constructionoperation, the outdoor environment of the construction process makes the off-linepath planning especially important and requires the planning function to beequipped with a capability for constant correction and modification. AsKunigahalli et al. (1993) argued, the potential of automation in construction can beincreased by providing motion planning capabilities whether with complete or withincomplete information, challenging the unstructured construction environment.
Architecture of the Graphic-based Off-line Path Planning
A graphic-based path planner is built in a graphic environment of CAD forefficient telerobotic operation of construction manipulators (Fig. 1). The CADintegration allows a user-interface for determining the trajectory. The trajectory isdrawn in a coordinate system developed for the working environment. Thegeometric configuration of the trajectory is depicted using a Bezier curve. Thepath from the starting position to the goal position can be divided into a number ofvia points. Using inverse kinematics, the planner calculates joint angles that arenecessary to automatically move the manipulator to the via point. The capabilityof collision avoidance is added to provide a real-time manipulator control. Keyaspects of the main architecture are described in detail as the following:
Structure of Coordinate Frame: A coordinate system is structured inorder to keep track of the manipulator position in relation to the world coordinateof the bridge deck. The overall coordinate system is made up of several localcoordinate systems that include bridge, crane, crane boom base, four crane boomjoints, two cameras, and target coordinates. The bridge coordinate serves as aworld coordinate and is used as a reference point by which the position andorientation of the end-effector can be indicated. Any point within the work spacearound the bridge structure can be indicated for systematic representation of themanipulation position.
Bezier Curves: The manipulator path is represented as a space curvetraced by the manipulator from a starting position to a desired destination. Themathematical function may be as simple as a straight line or so complicated thatthey require a high-order polynomial description. The Bezier function (Bezier1971) is used to describe the manipulator path for off-line path planning. Themathematical description of the curves provides trajectory data required for thetelerobotic operation of construction manipulators.
Inverse kinematics/ forward kinematics: The retrofitted Peeper crane hasspatial joints with four degrees of freedom. Given the via points, an inversekinematic algorithm is invoked to calculate the joint angles of the crane boomsections and to store the results in a trajectory table. Using the trajectory table, thedeployment of the manipulator is graphically simulated in a CAD environment.
Collision avoidance: In order to avoid collision with obstacles on themanipulator path, constraints should be given for the calculation of inversekinematics. The task is to solve the kinematics problem such that the crane boomcan avoid these objects. Ultrasonic sensors with wide beam angles are used fordetecting obstacles. They are mounted on the crane boom. Whenever the boomapproaches an obstacle, the sensors indicate the distance, and display warningsigns. The distance is then used for calculating the inverse kinematics in order tocreate the configuration of the crane boom.
This architecture provides a means to graphically simulate the motion ofthe crane boom on the planned path. Fig. 2shows a CAD model of the craneboom being animated following the generated path. During the simulation, humanoperators inspect the CAD model at various viewing angles using a dynamic viewfunction, and indicates the start and goal positions. Through visual inspection,they can understand the required configurations of the crane boom, as well asidentify the obstacle interference that may be encountered during field operations. During the simulation, operators can gain familiarity with the control requirements,and develop a control strategy. The procedure can be repeated until a feasibletrajectory is selected given a particular site condition. The trajectory is then usedfor automatically deploying manipulator booms under the bridge deck.
Field Experiment: Linkage to On-line Manipulator Control
The control architecture of off-line path planning is tested utilizing a large-scale construction manipulator. The Peeper crane and the testing facility at theNorth Carolina Department of Transportation (NC DOT) were used for thisexperiment. A virtual control model is designed to support the efficient human-machine interface during on-line manipulator control. The model is built such thatthe data from the off-line path planning can be directly used for advancing thecrane boom under the bridge deck. The control model takes advantage of a 3DCAD model that is superimposed onto live camera images. The model provides ameans for the operator to interact with the machine controller on a real-time basis. The procedure of implementing the control method was sequenced such that
1) selection of via points; 2) creation of a trajectory; 3) collision avoidance withobstacles; 4) graphic simulation of actual deployment; 5) evaluation of the results;6) selection of a feasible path; and 7) implementation of the accepted path for on-line manipulator control.
During the field experiment (Fig. 3), graphic overlay of the CAD modelonto live camera images was used to indicate the next via point to which themanipulator should move. Due to the inaccuracy of actual movement thetrajectory was constantly updated in order to adjust the originally planned path. Inthe control loop, the operator played a supervisory role that presides the overalltelerobotic operation. While looking at a TV monitor, the operator just hits acomputer key to advance the crane boom from one via point to next via point.
The typical task completion time for deploying the crane boom was rangedbetween 3 and 4 minutes. The duration approximately matches conventional joystick control for the Peeper crane control with a full line of sight. Considering thetelerobotic requirements with no line of sight during the Robotic BridgeMaintenance, however, the result demonstrates a capability for remote manipulatorcontrol. The end-tip position of the crane boom was controlled within one foot ofthe planned trajectory. The overall efficiency of the graphic-interfaced manipulatorcontrol is under study.
Summary
The unique characteristics of the construction environment necessitates acareful approach in implementing an automation paradigm to a constructionoperation. The graphic-based planner provides a basis for an efficient on-linemanipulator control of construction manipulators. Graphic simulation of themanipulator configuration provided operators with visual preview before actualoperation is executed. This enhanced operators' confidence in the manipulatorcontrol. The planner not only offers a chance to determine a desirable path forreaching a goal position, but also provides a means for operators to trainthemselves prior to the actual operation of the manipulator. The confidence inturn reduced trial and errors during actual manipulator control and enhancedcontrol performance. The study result showed that the integration of the humanoperator's decision capability and the machine controller's automation providesnew opportunities for efficient manipulator control in the constructionenvironment.
References
Craig, J. J. (1986). Introduction to Robotics: Mechanics & Control, Addison- Wesley Publishing Co., Reading, Massachusetts, 295-297.
Kunigahalli, R., Russell, J. S., and Skibniewski, M. J. (1993). "Motion planning for automated construction." Automation and Robotics in Construction X, Proc. 10th International Symposium on Automation and Robotics in Construction, Houston, TX, 407-413.
Schilling, R. J. (1990). Fundamentals of Robotics: Analysis & Control, Prentice- Hall, Inc. Englewood Cliffs, New Jersey
Brooks, R. A. (1983). "Solving the find-path problem by good representation of free space." IEEE Trans. Syst., Man, Cybern., SMC-13(3), 190-197.
Lozano-Perez, T. (1983). "Spatial planning: A configuration space approach," IEEE Trans. Computers, C(32), 108-110.
Lee, D. T., and Drysdale, R. L. (1981). "Generalized Voronoi diagrams in the plane." SIAM J. Compt., 10(1), 73-87.
