Local Planner Bench Benchmarking for Local Motion Planning Max Spahn Chadi Salmi Javier Alonso-Mora Abstract Local motion planning is a heavily researched topic

2025-04-27 0 0 625.03KB 4 页 10玖币

侵权投诉

Local Planner Bench: Benchmarking for Local Motion Planning

Max Spahn∗, Chadi Salmi∗, Javier Alonso-Mora∗

Abstract—Local motion planning is a heavily researched topic

in the ﬁeld of robotics with many promising algorithms being

published every year. However, it is difﬁcult and time-consuming

to compare different methods in the ﬁeld. In this paper, we

present localPlannerBench, a new benchmarking suite that

allows quick and seamless comparison between local motion

planning algorithms. The key focus of the project lies in the

extensibility of the environment and the simulation cases.

Out-of-the-box, localPlannerBench already supports many

simulation cases ranging from a simple 2D point mass to

full-ﬂedged 3D 7DoF manipulators, and it’s straightforward to

add your own custom robot using a URDF ﬁle. A post-processor

is built-in that can be extended with custom metrics and

plots. To integrate your own motion planner, simply create a

wrapper that derives from the provided base class. Ultimately

we aim to improve the reproducibility of local motion planning

algorithms and encourage standardized open-source comparison.

Local Motion Planning, Benchmark, Collision Avoidance

Code: github.com/tud-amr/localPlannerBench 1

I. INTRODUCTION

Robot motion planning in dynamic environments is fundamen-

tally different from motion generation in static environments

because initial plans must be constantly updated in the pres-

ence of unforeseen events. Algorithms that aim to adapt global

plans according to dynamic changes in the environment and

transfer them into actions at runtime fall into the category

of local motion planning or reactive motion planning. As

the number of applications (e.g. mobile robots, robotic arms,

mobile manipulators) is quite diverse, different methods have

been presented in the last years showing varying performance

depending on the scenarios. These methods can be roughly

divided into four categories: (a) geometric approaches, such as

the potential ﬁeld method [1], reciprocal collision avoidance

with velocity obstacles [2], Riemannian Motion Policies [3]

or optimization fabrics [4], [5], (b) optimization-based ap-

proaches, such as STOMP [6] or model predicitive control

[7], [8], (c) compositon of motion primitives [9], [10], and

(d) learning-based approaches [11], [12]. Due to a steady

increase in number of methods and a tendency to focus on

one of the above approaches, it becomes increasingly difﬁ-

cult and time-consuming to make proper comparisons among

methods. Additionally, new methods in the ﬁeld are often not

publicly available so comparisons require re-implementation

of these methods. Together with a general bias toward op-

timizing parameters of a proposed method, fair comparisons

become more challenging. To address these issues, we offer

localPlannerBench, a bench-marking tool to allow quick and

seamless comparison between different local motion planning

algorithms. Additionally, we promote extendibility to allow

users to modify localPlannerBench according to their setups.

1version 1.0.0: https://github.com/tud-amr/localPlannerBench/tree/v.1.0.0

First, we review brieﬂy other benchmarking software with

a similar purpose while laying out the differences to our

framework (Section II). Then, we state the scope of our

framework and introduce the approach. (Section III). We lay

out the general structure of localPlannerBench and explain

individual components in Section III-E. Lastly, we brieﬂy

present a use-case where two different planners are compared

(Section IV).

II. RELATED WORK

As we propose localPlannerBench for benchmarking local

motion planning algorithms, we brieﬂy revise existing bench-

marks in robotics to highlight differences. Sampling-based

motion planning for global motion planning was ﬁrst stan-

dardized in the Open Motion Planning Library (OMPL) [13].

There, many algorithms are implemented and accessible to

the user. Today, it is used as the backend planning library

for many robotics software [14], [15]. While the beneﬁt of

assembling different algorithms in a single library is, still

today, highlighted by OMPL, it is not a benchmark suite on its

own. Robowﬂex is a wrapper around various motion planning

libraries that can evaluate different global motion planning

algorithms in isolation [16]. It also includes standardized

motion planning problems to promote comparisons. The goal

behind localPlannerBench is to provide a similar framework

but instead of focusing on global motion planning methods,

which are most valuable in static environments, we focus

on dynamic environments and thus on local motion plan-

ning methods. Some benchmarking frameworks focus purely

on mobile robots and/or autonomous driving and provide

benchmarks for local motion planning methods in cluttered

and human-shared environments [17], [18]. With localPlan-

nerBench, we generalize this approach beyond mobile robots

by including robotic arms and mobile manipulators. To make

motion planning methods more reproducible and comparable,

all of the mentioned works play a central role and localPlan-

nerBench takes this role for local motion planning.

III. APPROACH

A. Scope

We deﬁne local motion planning as the problem of computing

actions for the robot at runtime, such that the sequence of

actions leads to progress in the fulﬁllment of the goal while

avoiding collisions with its environment. From this deﬁnition,

it follows that interactive motion planning, where the robot is

allowed to manipulate objects, is not considered. Moreover,

we consider collision avoidance the only safe way of navi-

gating through the environment. While the initial state can be

speciﬁed by a robot conﬁguration, the goal is a composition

of geometric constraints. Some examples that can be stated

arXiv:2210.06033v1 [cs.RO] 12 Oct 2022

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

LocalPlannerBench:BenchmarkingforLocalMotionPlanningMaxSpahn,ChadiSalmi,JavierAlonso-MoraAbstractLocalmotionplanningisaheavilyresearchedtopicintheeldofroboticswithmanypromisingalgorithmsbeingpublishedeveryyear.However,itisdifcultandtime-consumingtocomparedifferentmethodsintheeld.Inthispaper,w...

展开>> 收起<<

Local Planner Bench Benchmarking for Local Motion Planning Max Spahn Chadi Salmi Javier Alonso-Mora Abstract Local motion planning is a heavily researched topic.pdf

共4页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Local Planner Bench Benchmarking for Local Motion Planning Max Spahn Chadi Salmi Javier Alonso-Mora Abstract Local motion planning is a heavily researched topic

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: