CCQ: Efficient Local Planning Using Connection Collision Query

Abstract

We introduce a novel proximity query, called connection collision query (CCQ), and use it for efficient and exact local planning in sampling-based motion planners. Given two collision-free configurations, CCQ checks whether these configurations can be connected by a given continuous path that either lies completely in the free space or penetrates any obstacle by at most ε, a given threshold. Our approach is general, robust, and can handle different continuous path formulations. We have integrated the CCQ algorithm with sampling-based motion planners and can perform reliable local planning queries with little performance degradation, as compared to prior methods. Moreover, the CCQ-based exact local planner is about an order of magnitude faster than prior exact local planning algorithms.

Collision-free and smooth trajectory computation in cluttered environments

“…In order to determine the conservative time step, it needs to compute the minimal separation distance between the objects and uses it to estimate conservative motion bounds. Tang et al [Tang et al 2011] present algorithms to compute the motion bound for screw motion. Conservative advancement algorithm can be applied to non-convex models based on BVHs [Tang et al 2011] and can be further extended to articulated models [Zhang et al 2007].…”

“…Our SCD (spline collision detection) algorithm is based on conservative advancement (CA) [Zhang et al 2006;Tang et al 2011;Pan et al 2011]. Figure 4 shows an overview of the CA method.…”

“…Tang et al [Tang et al 2011] present the motion bounds for linear and screw motion in the configuration space. In our case, we need to find an appropriate motion bound for the spline motion.…”

“…As mentioned above, motion bound µ is an upper bound on the motion of A(t) projected onto the direction n. Previous work [Zhang et al 2006;Tang et al 2011;Pan et al 2011] define the motion bound as µ = max…”

“…Moreover, it cannot be applied to non-convex objects. To solve these difficulties, we use the swept sphere volume (SSV) hierarchy technique to accelerate motion bound computation and extend it to nonconvex objects [Tang et al 2011]. As a preprocess, we compute the SSV hierarchy for both objects.…”