Path planning of 3D solid objects using approximate solutions

Path planning of 3D solid objects has many practical applications. Sampling-based planners like Rapidly-exploring Random Tree (RRT) tackle the problem by the randomized sampling of the configuration space. The well-known issue of the sampling-based planners is the narrow passage problem. Narrow passages are small collision-free regions in the configuration space that are, due to their low volume, difficult to cover by the random samples, which prevents the sampling-based planners from finding a path leading through the passages. To increase the success rate of the planners, the search in the configuration space can be guided using an approximate solution. An approximate solution is found, for example, considering a smaller object (robot). The performance of this planning concept depends on the way of reducing the size of the object and the ability to find an initial solution. In this paper, we propose a novel technique to reduce the geometry of the object by a combination of triangulation and iterative removal of surface triangles. This technique is suitable for both convex and non-convex objects. In many real-world applications, narrow passages can also contain the initial or goal configurations. We thus propose the extension of the goal region to increase the probability of finding the initial approximate solution. Experiments have shown that the proposed modification outperforms the simple guiding using approximate solutions, as well as other related state-of-the-art planners.

I. Introduction

Path planning of 3D solid objects, where the task is to find a collision-free trajectory for an object between two positions, has many practical applications in robotics and industry, e.g., in bin-picking and data collection [26] . In virtual prototyping, path planning helps to determine a sequence of actions to assemble a product from its parts [1] , [12] . Computational biochemistry utilizes path planning to detect exit pathways from proteins [6] or to identify binding sites [3] .