TEB_DWA.rar_teb算法matlab资源-CSDN文库

共9个文件

pdf：8个

txt：1个

局部路径规划

需积分: 50 37 浏览量 2019-05-24 10:58:02 上传评论 4 收藏 3.28MB RAR 举报

资源推荐

资源详情

资源评论

收起资源包目录

TEB_DWA.rar （9个子文件）

DWA

The Dynamic Window Approach to Collision Avoidance.pdf 581KB

An Intelligent Predictive Controller.pdf 108KB

Guided Dynamic Window Approach to Collision.pdf 314KB

TEB

Planning of Multiple Robot Trajectories in Distinctive Topologies.pdf 372KB

Planning of Optimal Collision Avoidance.pdf 369KB

新建文本文档.txt 241B

Efficient Trajectory Optimization using a Sparse Model.pdf 414KB

Trajectory modification considering dynamic constraints.pdf 476KB

Integrated online trajectory planning and optimization in distinctive.pdf 1.33MB

Robotics and Autonomous Systems 88 (2017) 142–153

Contents lists available at ScienceDirect

Robotics and Autonomous Systems

journal homepage: www.elsevier.com/locate/robot

Integrated online trajectory planning and optimization in distinctive

topologies

Christoph Rösmann *, Frank Hoffmann, Torsten Bertram

Institute of Control Theory and Systems Engineering, Technical University of Dortmund, 44227 Dortmund, Germany

h i g h l i g h t s

• An integrated online trajectory optimization approach is proposed.

• Maintaining and optimization of admissible candidate trajectories of distinctive topologies in order to seek the overall best solution.

• An exploration strategy based on Voronoi diagrams provides the complete set of alternative trajectories.

• An alternative proposed sampling based strategy generates a sufficient subset for small to medium sized environments under limited computational

budgets.

• The Timed-Elastic-Band trajectory optimization method complies with non-holonomic kinematics of differential-drive and carlike robots.

a r t i c l e i n f o

Article history:

Received 31 January 2016

Accepted 10 November 2016

Available online 12 November 2016

Keywords:

Online trajectory optimization

Mobile robot motion planning

Distinctive topologies

Homology classes

Timed-Elastic-Band

Model predictive control

a b s t r a c t

This paper presents a novel integrated approach for efficient optimization based online trajectory

planning of topologically distinctive mobile robot trajectories. Online trajectory optimization deforms an

initial coarse path generated by a global planner by minimizing objectives such as path length, transition

time or control effort. Kinodynamic motion properties of mobile robots and clearance from obstacles

impose additional equality and inequality constraints on the trajectory optimization. Local planners

account for efficiency by restricting the search space to locally optimal solutions only. However, the

objective function is usually non-convex as the presence of obstacles generates multiple distinctive local

optima.

The proposed method maintains and simultaneously optimizes a subset of admissible candidate

trajectories of distinctive topologies and thus seeking the overall best candidate among the set of

alternative local solutions. Time-optimal trajectories for differential-drive and carlike robots are obtained

efficiently by adopting the Timed-Elastic-Band approach for the underlying trajectory optimization prob-

lem. The investigation of various example scenarios and a comparative analysis with conventional local

planners confirm the advantages ofintegratedexploration,maintenance and optimization of topologically

distinctive trajectories.

1. Introduction

In the context of service robotics and autonomous transporta-

tion systems, mobile robots are required to navigate in highly

dynamic environments while accomplishing complex tasks. On

this occasion, one of the fundamental challenges in mobile robotics

is concerned with the development of universally applicable mo-

tion planning strategies. Online planning is preferred over offline

approaches since they immediately respond to changes in the

environment or perturbations of the robot motion at runtime. The

well known elastic band approach [1] locally deforms a path online.

Corresponding author.

E-mail address: christoph.roesmann@tu-dortmund.de (C. Rösmann).

Predefined internal forces contract the path while external forces

maintain a separation from obstacles. An alternative established

path planning approach based on an optimization technique is

presented in [2]. However, conventional path planning does not

explicitly incorporate temporal and (kino-)dynamic aspects of mo-

tion, therefore ignoring constraints imposed by kinematic or dy-

namic motion models with bounded velocities and accelerations.

Kurniawati et al. extend the elastic band approach to the online

deformation of trajectories rather than paths [3]. The approach

consists of two stages that at first repel discrete trajectory points

from obstacles and secondly enforce connectedness w.r.t. a dy-

namic motion model. Delsart et al. combine both stages into a

single operation [4].

http://dx.doi.org/10.1016/j.robot.2016.11.007

C. Rösmann et al. / Robotics and Autonomous Systems 88 (2017) 142–153 143

Trajectory optimization usually attempts to minimize either

control effort, control error or the transition time between start

and goal pose. However, the computational burden required to

find an optimal solution is high, which limits its direct online

integration with feedback control. Due to this observation many

researchers are focusing on obtaining solutions or even approx-

imations of the underlying trajectory optimization problem ef-

ficiently. The dynamic window approach (DWA) constitutes a

widely applied method for mobile robot navigation [5]. Simulated

trajectories are sampled repeatedly from a velocity search space

restricted by a set of feasible velocities. A cost function evaluates

each candidate trajectory w.r.t. the remaining distance to the goal,

the forward velocity and separation from obstacles. The lowest-

cost solution is selected for controlling the robot. The approach

accounts for efficiency by restricting the feasible set of valid trajec-

tories to a subset of sampled candidates with segments of constant

velocity leading to merely suboptimal trajectories. Lau et al. [6]

and Sprunk et al. [7] optimize trajectories represented by splines

continuously according to kinodynamic constraints of the robot.

An online planning algorithm that relies on a covariant gradient

descent method is presented in [8]. In a previous work the authors

present a further extension to the elastic band called Timed-Elastic-

Band (TEB) approach [9,10]. The TEB efficiently optimizes the robot

trajectory w.r.t. (kino-)dynamic constraints and non-holonomic

kinematics while explicitly incorporating temporal information

in order to reach the goal pose in minimal time. The approach

accounts for efficiency by exploiting the sparsity structure of the

underlying problem formulation. It has been generalized to an effi-

cient time-optimal model predictive control approach for dynamic

systems [11].

In practice, due to limited computational resources online opti-

mization is usually performed using local optimization techniques

for which the discovery of the global optimal trajectory is not

guaranteed. Especially in mobile robot navigation local minima

often emerge due to the presence of obstacles. Kalakrishnan et al.

apply a stochastic descent method to partially overcome these lim-

itations [12]. However, the approach requires extensive sampling

of trajectories in order to estimate the true gradient w.r.t. the global

optimum. A two stage local optimization approach that generates a

set of alternative, topologically distinctive trajectories is presented

by Kuderer et al. [13]. The proposed method extracts multiple

candidate trajectories from a modified Voronoi diagram that often

coincide with the local minima of the optimization problem. Paths

that belong to the same equivalence class are grouped by an equiv-

alence relation based on the winding number at each obstacle. The

paper at hand pursues a related approach for filtering distinctive

candidate trajectories in the context of online trajectory planning.

It mainly differs from [13] w.r.t. sampling strategy, equivalence

relation and the trajectory optimization technique.

The idea of exploring topologically distinctive paths and tra-

jectories is not novel. However, past approaches mainly focus on

global offline path and trajectory planning. Probabilistic roadmap

(PRM) methods that operate with different equivalence relations

are presented in [14] and [15]. The proposed algorithms are in

principle able to identify complicated paths in large, complex

environments but rely upon an algorithmic rather than closed form

computation of the equivalence relations. Our approach employs a

computationally more efficient sampling strategy and equivalence

relation. It is intended to operate in a local subregion of the en-

vironment and is thus suited for online trajectory optimization.

Obviously, this does not replace the need for global planning in

large environments. Knepper et al. propose a local planner based

on path sampling and the Hausdorff metric as equivalence rela-

tion [16]. The planner performs a discrete path selection rather

than deforming continuous trajectories. An equivalence relation

originating from the field of complex analysis is presented by

Fig. 1. TEB trajectory representation with n = 3 poses.

Bhattacharya et al. [17]. The closed form solution motivates its

application within our approach for trajectory filtering. Pokorny

et al. presents a graph free sampling based approach based on

filtrations of simplicial complexes [18].

This contribution presents an integrated online trajectory plan-

ning approach that combines the exploration and simultaneous

optimization of multiple admissible topologically distinctive tra-

jectories during runtime. As an extended version of [19] the online

exploration strategy is fully integrated with the TEB approach

providing the underlying trajectory optimization. Hence the TEB

approach is entirely reformulated w.r.t. its original description [9]

and further extended to a more generic obstacle representation,

supporting forward and backward robot motion and to accomplish

navigation tasks for carlike robots beyond the differential-drive

robots considered in the original proposal. The combined and in-

tegrated approach preserves the locally optimal trajectory of each

equivalence class to facilitate warm starting from previous solu-

tions. Candidate trajectories are sampled repeatedly and evolve

not only over space but also time which significantly reduces the

number of samples. The sample based approach is compared with

the systematic generation of waypoints from Voronoi diagrams re-

garding completeness of candidate trajectories and computational

efficiency. The presented approach is available as open-source C++

code and integrated into ROS [20].

The paper is organized as follows: Section 2 formulates the

local TEB optimization approach followed by the exploration of

trajectories in distinctive topologies in Section 3. The overall in-

tegration of exploration, planning and maintenance of trajectories

is described in Section 4. Section 5 demonstrates and evaluates

results obtained from realistic simulations with a differential drive

and a carlike mobile robot. The integrated planning approach is

compared with the original locally operating TEB and the DWA as

mentioned above. Finally, Section 6 summarizes the results and

provides an outlook on future work.

2. Timed-elastic-band approach

This section describes the TEB approach which performs the

actual trajectory optimization in the overall planning task. The ap-

proach is entirely reformulated and extended w.r.t the formulation

originally presented in [9].

2.1. Definition and representation

Let s

= [x

, y

, β

]

⊺

∈ R

×S

denote a robot pose at a discrete

point in time k. x

∈ R and y

∈ R represent the planar position

and β

∈ S

the orientation of the robot. A discretized trajectory is

described in terms of a sequence S = {s

|k = 1, 2, . . . , n} of robot

poses. The TEB augments the trajectory representation with strictly

positive time intervals ∆T

∈ R

, k = 1, 2, . . . , n − 1. Each time

interval denotes the time that the robot requires to transit from the

pose s

to its subsequent pose s

k+1

. An exemplary trajectory with

three poses is depicted in Fig. 1. Poses and time intervals are joined

into the parameter vector b ∈ B:

b = [s

, ∆T

, s

, ∆T

, s

, . . . , ∆T

n−1

, s

]

⊺

(1)

144 C. Rösmann et al. / Robotics and Autonomous Systems 88 (2017) 142–153

Fig. 2. Geometric interpretation of the non-holonomic constraint.

2.2. Open-loop optimization problem

The TEB open-loop optimization task is to find controls in

order to transit the robot from an initial pose s

to a final pose

in minimal time while satisfying kinodynamic constraints and

maintaining a safe separation from obstacles. The overall task is

formulated as a nonlinear program:

∗

(b) = min

n−1



k=1

∆T

(2)

subject to

= s

, s

= s

, ∆T

> 0

k+1

, s

) = 0,

k+1

, s

) ≥ 0,

k+1

, s

, ∆T

) ≥ 0, (k = 1, 2, . . . , n − 1)

k+2

, s

k+1

, s

, ∆T

k+1

, ∆T

) ≥ 0, (k = 2, 3, . . . , n − 2)

, s

, ∆T

) ≥ 0, α

, s

n−1

, ∆T

n−1

) ≥ 0.

According to the definition of the trajectory (1), the total transition

time is approximated by T ≈



n−1

k=1

∆T

. The minimization of T is

ill-posed since individual ∆T

are unconstrained. Instead, temporal

homogeneity is achieved by minimizing the sum of squared ∆T

which enforces uniform time intervals ∆T

. Initial s

and

final pose s

are constrained by s

and s

respectively. Kinematic

constraints between two consecutive poses s

and s

k+1

are in-

corporated by equality constraints h

k+1

, s

). Furthermore, for

carlike robots a minimum turning radius must be satisfied which

is captured by r

k+1

, s

). The robot’s velocity and acceleration are

limited by inequalities ν

(·) and α

(·) respectively. The inequality

) sustains a minimum separation from obstacles. Each term is

described in more detail in the following paragraphs.

Non-holonomic kinematics.

The equality constraint ν

(·) enforces compliance with the kine-

matic constraints of the mobile robot. This paper investigates in

particular differential drive and carlike robots with only two local

degrees of freedom, thus the robot can only execute a smooth path

that is composed of linear and arc segments. According to [9], the

non-holonomic constraint is defined by a geometric interpretation.

Two consecutive poses s

and s

k+1

are required to be located on a

common arc of constant curvature as shown in Fig. 2. The angle ϑ

between pose s

and the direction d

k,k+1

= [x

k+1

−x

, y

k+1

−y

, 0]

⊺

has to be equal to the corresponding angle ϑ

k+1

at the consecutive

pose s

k+1

= ϑ

k+1

(3)



cos(β

)

sin(β

)



×d

k,k+1

= d

k,k+1



cos(β

k+1

)

sin(β

k+1

)



. (4)

The resulting equality constraint is given by:

k+1

, s

) =



cos(β

)

sin(β

)





cos(β

k+1

)

sin(β

k+1

)



×d

k,k+1

. (5)

Equation (5) is suitable for differential drive robots that are

able to rotate in place. For carlike robots and Ackermann drives

respectively, the robot motion is further restricted by a minimal

turning radius between consecutive poses. The absolute turning

radius

r is given by:

r(s

k+1

, s

) =



≈

∥d

k,k+1

∥

k+1

− β

. (6)

In order to satisfy a minimal turning radius of carlike robots, an

inequality constraint with r

k+1

, s

) =

r(·) − r

min

is introduced.

min

denotes the lower bound on the turning radius. For differential

drive robots it becomes r

min

= 0.

Limited velocity and acceleration.

Limiting the velocity is crucial for the success in practical appli-

cations. Without any bound, the time-optimal solution according

to (2) implies ∆T

→ 0, ∀k as v

→ ∞. For the sake of simplicity,

translational and rotational velocities at each pose v

and ω

(e.g. in

the center of the robot) are considered rather than individual wheel

velocities. This simplification is sufficient for most practical appli-

cations. However, the approach allows the limitation of arbitrary

velocities that can be expressed in terms of v

, ω

and dedicated

robot design parameters. Translational and rotational velocities are

approximated using finite differences according to the Euclidean

resp. angular distance between two consecutive poses s

and s

k+1

= ∆T

−1

∥[x

k+1

− x

, y

k+1

− y

]

⊺

∥γ (s

, s

k+1

) (7)

= ∆T

−1

(β

k+1

− β

). (8)

Note, subtracting and adding angles in domain S

requires a

normalization in practical implementations. γ (s

, s

k+1

) denotes

a function that extracts the sign of the translational veloc-

ity, whether the robot moves forwards or backwards. For non-

holonomic robots that are restricted to transitions along linear and

arc segments respectively, the sign of the projection of orientation

vector q

= [cos β

, sin β

, 0]

⊺

onto the distance vector d

k,k+1

utilized:

γ (s

, s

k+1

) = sign (⟨q

, d

k,k+1

⟩). (9)

Hereby, operator ⟨·, ·⟩ applies the scalar product. Since many com-

mon optimization algorithms are not suited for non-smooth func-

tions such as (9), a sigmoidal approximation maps the projection

to the interval [−1, 1]. Our implementation employs:

γ (s

, s

k+1

) ≈

κ⟨q

, d

k,k+1

⟩

1 + |κ⟨q

, d

k,k+1

⟩|

. (10)

Variable κ ∈ R

denotes a scaling factor that changes the slope

(e.g. κ = 10

). Eq. (9) and (10) result in vanishing and incorrect

velocities at ⟨q

, d

k,k+1

⟩ = 0 or ⟨q

, d

k,k+1

⟩ ≈ 0 respectively.

Figuratively, this occurs if pose s

k+1

is placed orthogonal to pose

. However, such configuration is not part of the feasible set given

by the non-holonomic constraint as mentioned before and for the

special case in which position parts of both poses coincide it is

k,k+1

= 0 H⇒ v

= 0.

Limiting velocities to ±v

max

and ±ω

max

is obtained by the in-

equality constraint ν

k+1

, s

, ∆T

) = [v

max

− |v

|, ω

max

− |ω

⊺

The equations might be adapted in cases in which the bound

on negative velocities should differ, which is neglected here for

simplicity, but which is often preferred in practical applications.

C. Rösmann et al. / Robotics and Autonomous Systems 88 (2017) 142–153 145

A similar procedure is applied for limiting translational and rota-

tional accelerations a

and ˙ω

respectively. In particular for a

it is:

2(v

k+1

− v

)

∆T

+ ∆T

k+1

. (11)

For the sake of clarity, s

k+2

, s

k+1

and s

are substituted by their

related velocities (7). The limitation of the acceleration is obtained

by inequality α

k+2

, s

k+1

, s

, ∆T

k+1

, ∆T

) = [a

max

− |a

|, ˙ω

max

−

| ˙ω

⊺

. Special cases occur at k = 1 and k = n − 1 for which v

and

n−1

are substituted by desired start and final velocities (v

, ω

) and

, ω

) respectively.

Obstacle avoidance.

The robot is supposed to reach the goal without any collision

with obstacles. The approach is suited for a broad range of obstacle

representations. The underlying optimization method is feasible

and computationally efficient under the assumption that the min-

imal Euclidean distance between the pose s

and the set of points

on the obstacle perimeter is described by a continuous function.

This paper considers obstacle representations in terms of set of

points, circles, lines and polygons. An obstacle is represented as a

simply-connected region in R

and is denoted as O. In the presence

of R obstacles O

, l = 1, 2, . . . , R, the subscript l is added. Let

ρ(s

, O) : R

×S

×O → R denote the minimal Euclidean distance

between obstacle O and the pose s

. The orientation part β

of s

can be neglected in case of a circular shape robot. The inequality

constraint maintains a minimum separation ρ

min

between all ob-

stacles and pose s

according to:

) = [ρ(s

, O

), ρ(s

, O

), . . . , ρ(s

, O

)]

⊺

− [ρ

min

, ρ

min

, . . . , ρ

min

]

⊺

. (12)

The equality constraint (12) in itself restricts the feasible set of

robot positions (x

, y

) to {R

\ (



l=1

)} in which

denotes

the obstacle region O

inflated by ρ

min

. Obviously, this set is non-

convex such that the corresponding nonlinear program (2) exhibits

multiple local minima. The occurrence of local minima caused by

the presence of obstacles is a common phenomenon in trajectory

and path planning. The number of obstacles l has a significant

influence on the number of local minima. Due to limited com-

putational resources local optimization methods are preferred for

online optimization, therefore the solution strongly depends on

the globally planned trajectory b which serves as an initial solution

for the local optimizer. Note, for most practical applications the

number of distance calculations required during the optimization

might be reduced such that each obstacle O

only effects a subset

of nearby poses, rather than the entire sequence of poses s

, ∀k.

The association between obstacles and nearby poses is updated

between subsequent sampling intervals to refine suboptimal so-

lutions during runtime.

2.3. Approximative least-squares optimization

Solving nonlinear programs with hard constraints is computa-

tional expensive. Therefore, improving the efficiency of fast on-

line solvers has become an important research topic over the

last decade. The TEB approach further investigates the applica-

tion of unconstrained optimization techniques since they are well

studied and mature implementations in open-source packages

are widely available. Unconstrained optimization avoids Lagrange

resp. Karush–Kuhn–Tucker multipliers (dual variables) causing the

dimension of the Hessian matrices to become identical with the

number of primal variables in b.

The exact nonlinear program (2) is transformed into an ap-

proximative nonlinear least-squares optimization problem that

is solved efficiently as the solver approximates the Hessian by

first order derivatives and exploits the sparsity pattern of the

problem. Constraints are incorporated into the objective function

as additional penalty terms. Since the unconstrained objective

function is composed of squared nonlinear terms only, quadratic

penalty functions are applied according to [21]. Note, that other

unconstrained approximations such as log-barrier, augmented La-

grangian or exact penalty methods [22] exist, which however

contain non-squared terms and therefore are not applicable to

unconstrained least-squares solvers.

In the following arguments of constraints are omitted for better

readability. The equality constraint h is expressed in terms of a

quadratic penalty with a scalar weight σ

and identity I by:

φ(h

, σ

) = σ

= σ

∥h

∥

. (13)

Inequalities are approximated by weighted one-sided quadratic

penalties:

χ(ν

, σ

) = σ

∥ min{0, ν

}∥

. (14)

The min-operator is applied row-wise. Further inequalities α

and

are approximated in a similar fashion. Initial and final con-

straint, s

and s

respectively, are eliminated by substitution and

are therefore not subject to the optimization. ∆T

> 0 is implicitly

incorporated by the difference quotients in (7) and (8) since the

quotients diverge for ∆T

→ 0 (initial ∆T

must be positive).

The overall unconstrained optimization problem with objective

function

V (b) that approximates (2) is given by:

∗

= arg min

B\{s

}

V (b) (15)

V (b) =

n−1



k=1



∆T

+ φ(h

, σ

) + χ(r

, σ

) + · · ·

+ χ(ν

, σ

) + χ(o

, σ

) + χ(α

, σ

)



+ χ (α

, σ

). (16)

Variable b

∗

denotes the optimal parameter vector. From the theory

of quadratic penalties [21] it is well known that b

∗

only coincides

with the actual minimizer of the nonlinear program (2) in case

all weights tend toward infinity σ → ∞. Unfortunately, large

weights introduce ill-conditioned characteristics of the problem

such that the underlying solver does not converge properly due

to inadequate step sizes. The TEB approach abandons the true

minimizer in favor of a suboptimal but computationally more

efficiently obtained solution with user defined weights. For small

to medium sized cluttered environments within the robots local

field of perception our experiments reveals that unit weights of

1 provide a reasonably point of departure, except for the weight

associated with the equality constraint of the non-holonomic

kinematics which should be chosen a few magnitudes higher

(≈ 1000). This setting is recommended as the penalty function (13)

is small compared to the inequality constraints and a near-perfect

compliance of the trajectory with the robot kinematics is crucial.

The transformation of the non-convex signed velocity con-

straint (10) onto a quadratic penalty term introduces local minima

(see Fig. 3a). However, the Fig. 3b reveals that these local minima

are canceled in the overall objective function by the conflicting

non-holonomic-constraint (5).

2.4. Solution of the approximative optimization problem

The literature proposes an abundance of algorithms for solving

nonlinear least-squares problems such as (15). Popular solvers are

Gauss–Newton, Dogleg or Levenberg–Marquardt (LM) algorithm

[21]. The TEB approach utilizes LM due to its proper balance

between robustness and efficiency. LM constitutes a trust-region

strategy that only accepts step sizes that decrease the overall

cost. The optimization problem is regularized implicitly in case it

评论收藏

内容反馈

zhangstar3

粉丝: 0
资源: 3

TEB_DWA.rar

TB_Day_Planner

ist的matlab代码-Apollo_13:使用teb-planner在类似汽车的机器人上实现ROS导航堆栈。实施车道检测。实施自主探索

对ROS局部运动规划器Teb的理解

ros_teb路径规划开发包

teb_local_planner_tutorials-kinetic_teb完整机器人_ROS_TEB路径_

teb_local_planner.zip

key_led.rar_key_led_测按键

CAD常用字库（275种字库）

C++ Convert 多种字符转换函数

jprofiler_windowsx64_itmop.com.zip

机器人局部路径规划算法——VFH系列论文

车辆导航中的快速路径规划算法

ros中的五种2D SLAM算法.pdf

路径规划与壁障算法 经典讲义

凸优化与A*算法结合的路径避障算法

teb_local_planner:考虑基于时间弹性带（ROS软件包）的移动机器人独特拓扑的最优轨迹规划器

test_sam.zip_Polar码_Polar译码matlab_polar SNR_polar code_sc译码

Autonom_2wr_and_4Dof_Robot_arm:使用teb_local_planner和robot_localization（GPS，IMU，Odom）套件的自动两轮车。 附在这辆车上的4-Dof机械臂。（Moveit，Python，Ros_control，ikfast_plugin）

frenet_planner:最优轨迹生成

wheelchair_simulation_navigation

局部RRT路径规划matlab代码-Planning-and-Control-for-Nonholonomic-Robot-Among-Obs

CAD字体，2000多种字体，常用各种cad图字体

视频图matlab代码-create_path_planner:MatlabiRobot创建模拟器的路径规划器

adda.rar_ADDA

mcl_pi_gazebo：使用粒子相交算法的多机器人协作定位

TEB-S7011新品速递.pdf

Graph-Search

Qt 5实现串口调试助手 （源工程文件、0积分下载）

【SystemVerilog】路科验证V2学习笔记（全600页）.pdf

最新资源

路径规划与壁障算法经典讲义

Autonom_2wr_and_4Dof_Robot_arm:使用teb_local_planner和robot_localization（GPS，IMU，Odom）套件的自动两轮车。附在这辆车上的4-Dof机械臂。（Moveit，Python，Ros_control，ikfast_plugin）

Qt 5实现串口调试助手（源工程文件、0积分下载）