Design and Evaluation of Motion Planners for Quadrotors in Environments with Varying Complexities

TL;DR

This work develops a procedure to construct parametrized environments, metrics that characterize the difficulty of motion planning in these environments, and an open-source software stack that can be used to combine a wide variety of two-stage planners seamlessly.

Abstract

Motion planning techniques for quadrotors have advanced significantly over the past decade. Most successful planners have two stages: a front-end that determines a path that incorporates geometric (or kinematic or input) constraints and specifies the homotopy class of the trajectory, and a back-end that optimizes this path to respect dynamics and input constraints. While there are many different choices for each stage, the eventual performance depends critically not only on these choices, but also on the environment. Given a new environment, it is difficult to decide a priori how one should design a motion planner. In this work, we develop (i) a procedure to construct parametrized environments, (ii) metrics that characterize the difficulty of motion planning in these environments, and (iii) an open-source software stack that can be used to combine a wide variety of two-stage planners seamlessly. We perform experiments in simulations and a real platform. We find, somewhat conveniently, that geometric front-ends are sufficient for environments with varying complexities if combined with dynamics-aware backends. The metrics we designed faithfully capture the planning difficulty in a given environment. All code is available at https://github.com/KumarRobotics/kr_mp_design

Figures (6)

• Figure 1: Overall trajectory evaluation for planning methods in different environments. The top figure demonstrates different front-end initial paths in a maze map and the bottom one shows the comparison of different back-end trajectories (thick lines) in a parameterized 3-D obstacle map. The blue polytopes are safe flight corridors.
• Figure 2: The architecture overview of the evaluation pipeline. The inputs for planning are the start state, goal tasks, and the standard environments from different sensors. The trajectory planning is modularized into the front-end and back-end stages, and divided in terms of its planning strategies and problem formulation.
• Figure 3: Examples of real maps (top), maze maps (middle), and obstacle maps (bottom) with different ECS values.
• Figure 4: The visualization of the distribution of different maps. The left figure demonstrates the ECI values of maze maps, obstacle maps, and real maps. The right three figures show how many planners (out of the 10 we consider) succeeded in generating a safe trajectory .
• Figure 5: GMM clusters (n = 2) of real maps from M3ED Chaney_2023_CVPR and STPLS3D Chen_2022_BMVC . There is a distinct cluster (black) with similar density and clutter indices values around $[0.07, 0.04]$. Random samples from each cluster are shown.