Path Planning of Construction Manipulators Using Genetic Algorithms

This paper presents the work done to investigate the potential of applying Genetic Algorithms for path planning of construction manipulators. Construction manipulators can take the shape of specialized equipment .such as the pipe manipulator or common equipment such as cranes. The need for automating the path of construction manipulators has been identified in a number of past studies. Path planning is a well-established area in robotics research, work in this area focuses on identifying new algorithms and techniques to improve the accuracy and efficiency of the automated path planning process. The application of Genetic Algorithms for robotic path planning is ca relatively new development. This paper reviews different construction situations in which path planning is required and identifies the factors influencing path selection in construction situations. Genetic Algorithm models to find satisficing paths are proposed. The GA model and initial results for a simple test problem using a 2 DOF planar manipulator are also presented. 1.0 Introduction This paper presents the work done to investigate the potential of applying Genetic Algorithms for path planning of construction manipulators. Construction manipulators can take the shape of specialized equipment such as the pipe manipulator or common equipment such as cranes. In all cases, manual path planning is a cumbersome process. Tasks requiring the representation of space, modeling the kinematics of the equipment and checking for interference are not suited for manual analysis. As a result, the productivity of detailed operations planning becomes very low. The automation of the path planning task would reduce the chances of human error, increase planning productivity and allow the planners to evaluate alternate plans and thereby arrive at a better plan. Path Planning is a well-established area in Robotics. Numerous techniques have been formulated and tested. These techniques are classified based on criteria such as dimension of space, mobility of manipulator and obstacles, representation of space and nature of information gathering. A description of the basic issues involved in path-planning and concepts of conventional techniques are presented effectively by Latombe [Latombe 19911 and Schwartz [Schwartz et at. 19881. The application of Genetic Algorithms for path planning is relatively new. Genetic Algorithms are optimization techniques based on the mechanics of natural selection and genetics. In the path planning domain, they have been used as a solution tool to solve the inverse kinematics problem as well as a technique to determine optimal paths through real space and configuration space. As Genetic Algorithm implementations can be paralleled easily, it offers it potential to determine near-optimal paths through complex spaces in reasonable time. A fundamental coverage of genetic algorithms have been presented by Goldberg [Goldberg 1995], Applications of conventional path-planning techniques to construction manipulators have met with limited success. Assumptions on the geometry of obstacles and motion of manipulators have to be made in order to obtain reasonable performance. These assumptions tend to negate the practical utility of the system. Further, there are differences in the criteria used to assess the suitability of a path in conventional robotics and construction situations. Although the current scope of this work is limited to off-line planning, the results from the system are expected in reasonable time. Recent work utilizing the configuration space representation with a heuristic search algorithm required several hours of execution time to determine a path for a 3 DOF crane within a realistic model of lift area [Reddy 1998]. Thus when situations involving multiple manipulators or more complex workspaces are encountered, the performance of the system will be unacceptable. Based on this experience, it was decided to investigate alternate approaches such as Genetic Algorithms for path planning of construction manipulators. This paper consists of six sections . A brief review of the past work in the areas of path planning in construction and path planning using genetic algorithms are presented in the next section. The third section compares the factors influencing path optimization of construction manipulators with conventional robotics. Genetic Algorithm models incorporating these factors are proposed and