ROBUST MULTISENSOR FRAMEWORK FOR MOBILE ROBOT NAVIGATION IN GNSS-DENIED ENVIRONMENTS

1
A N N A
U N I V E R S I T Y
ROBUST MULTISENSOR FRAMEWORK FOR
MOBILE ROBOT NAVIGATION
IN GNSS-DENIED ENVIRONMENTS
J. Steffi Keran Rani
(Reg. No. 2015225022)
Under the Supervision of
Dr.M.Deivamani

2
2
ABSTRACT
 The proposed work presents an algorithm that detects and avoids
both static and dynamic obstacles that lie in the path of the mobile
robot, be it indoor or outdoor.
 The proposed system allows the range of the obstacle detection to be
modified as per demand.
 The proposed system presents an RRT (Rapidly-exploring Random
Trees) -based path planner to find the shortest path to the goal.
 In this context, the presented work introduces an efficient, reliable
and scalable multi-sensor fusion system for autonomous mobile robot
navigation in GNSS(Global Navigation Satellite System)-denied
environments.
4-Jun-21
J. STEFFI KERAN RANI 2015225022

4
4
LITERATURE SURVEY
4-Jun-21

5
5
LITERATURE SURVEY – PATH PLANNING
4-Jun-21

6
6
LITERATURE SURVEY – RRT ALGORITHM
4-Jun-21

7
7
 Inability to identify obstructions that are greater in size
than half of the image size. Examples: wall, door etc.
 Poor angular resolution and specular reflections.
 Higher memory requirement and non-optimality.
 Jagged and non-continuous paths towards the goal.
 Increased computational load and runtime.
 Precise only for known or stable environments.
 Requires large update and iteration time.
 Reduced reliability over long period and increased
particle degradation.
PROBLEMS IDENTIFIED IN EXISTING SYSTEMS
4-Jun-21

PROBLEM STATEMENT
PHASE 1:
 The proposed work aims to overcome the landmark ambiguity problems encountered during
the robot navigation.
 The fused outcome of various sensory data assures accurate self-initialization and
localization of the robot by reducing erroneous estimates.
PHASE 2:
 The objective of the work is to implement a novel vision-based obstacle
detection in dynamic environments.
 The framework uses the resultant model from the phase-1 to accurately
integrate the obstacle locations onto the environment model.
 The shortest, collision-free and smooth path is calculated based on Rapidly
exploring Random Tree (RRT) algorithm.
 The overall architecture aims to achieve optimization in multi sensor system for
localization and navigation of robots.
9
4-Jun-21

11
11
PHASE 2 – HIGH LEVEL ARCHITECTURE
4-Jun-21

12
12
 The proposed work presents a novel path planning technique for the mobile robot navigation in
environments with static and dynamic obstacles.
 The proposed system employs Rapidly exploring Random Trees (RRT) as its basic algorithm.
 The system employs a vision based obstacle detection, followed by the computation of an optimal,
feasible and collision-free path to the destination.
 The building blocks of the Phase-2 project are:
1. Environment Perception
2. Obstacle recognition
3. Environment modelling
4. Integration of the obstacle and model
5. Optimized Path Planning
6. Trajectory smoothing and filtering
PROPOSED SYSTEM
4-Jun-21

13
13
PHASE 2
 Sensor fusion of camera and odometry data
 Accurate pose estimation
 Accurate Localization in known environment
 Landmark based navigation
 Reduced Error rate
 Performance analysis
 Obstacle detection
 Collision Avoidance
 Optimized Path Planning
 Trajectory filtering and smoothing
 Performance comparison with other state-of-the-art
SLAM algorithms
 Metrics computation
 Hardware implementation of Obstacle avoidance
DELIVERABLES
PHASE 1
4-Jun-21

14
14
NAME Cognitive navigation dataset
SCENARIO Democritus University of Thrace
URL http://robotics.pme.duth.gr/kostavelis/Dataset.html#10
ROBOT MAGGIE (Mobile Autonomous riGGed Indoors Exploratory) Robot.
J. STEFFI KERAN RANI 2015225022 4-Jun-21
DATASET
14

15
15
 Obstacle detection plays an important role in mobile
robots which operate in a highly diverse environments.
 The proposed system employs a vision-based obstacle
detection which can detect the presence of both static
and moving obstacles.
 Vision-based obstacle detection is superior to other
range-based sensors which suffer from specular
reflections and poor angular resolution.
 The proposed work involves detecting the objects that
differ in appearance from the ground and identifying
them as obstacles.
 Finally, if there are connected components in the
resultant image, it indicates the presence of the obstacle.
4-Jun-21
MODULE 1 : COMPUTER VISION-BASED OBSTACLE DETECTION

17
17
 This algorithm aims to detect the static and dynamic obstacles in both open and
confined environments.
Algorithm Computer Vision-based Obstacle Detection
Input : Camera image I, Reference background image W
Output: State S of obstacle flag
1 Read the images I and W
2 Resize the images to a standard size s
3 Convert I and W to grayscale
4 Rb ← removeBackground( I,W )
5 Convert Rb to grayscale
6 RbTh ← otsuThreshold( Rb )
7 RoI ← bwAreaOpen( RbTh )
8 Label ← bwLabel(RoI)
9 if Label < 1 then
10 Set the flag F
11 return State St of the flag F
4-Jun-21

4
5
6
7
1 2 3
OUTDOOR
4-Jun-21
19

1 2 3
4
5
6
7
INDOOR
4-Jun-21
20

1 2 3
4
5
6
7
INDOOR
4-Jun-21
21

22
22
MODULE 2- PATH PLANNING AND
OBSTACLE AVOIDANCE

 The proposed system employs Rapidly-exploring Random Tree (RRT) algorithm as the principal technique
behind the path planning of robot. A two-step procedure of path planning is summarized as follows.
• Step 1: The first step is environment perception and modelling, usually using a grid map (with occupancy
probability)
• Step 2: Then path planning algorithm is employed to find the best path according to the cost function, with
the ability to achieve both time efficiency and cost minimum.
 An RRT is iteratively expanded by applying control inputs that drive the system slightly toward random points,
as opposed to requiring point-to-point convergence, as in the probabilistic roadmap approach.
 RRT-based path planning has many advantages like:
 Environmental coverage
 2D or 3D search space
 Obstacle Avoidance
 Auto-connect to Goal
 Path smoothing
MODULE 2 : RAPIDLY EXPLORING RANDOM TREE (RRT) PATH PLANNER
4-Jun-21
23

24
RRT-BASED PATH PLANNING
Path
Collision
Detection
Collided Branch
Deletion
Deformation
Resampling &
Extension
Node Deletion
Smoothing

RRT EXTENSION PROCEDURE 25

RRT-BASED PATH PLANNING 26
rrtPathPlanning.m

27
27
 RRT grows a randomized tree during search. It terminates once a state close to
the goal is expanded.
RRT* Planner
Algorithm RRT Planner
Input : TreeMax, SeedsPerAxis, wallCount
Output: RRT graph Graph
1 Check inputs
2 Initial plotting and environmental setup
3 Continue search while the number of steps is less than TreeMax
4 Generate a new point
5 Find Nearest neighbour and connect to it
6 Draw the Output

28
28
RRT* Generation
Algorithm RRT
Input : Node Start, K, Node Goal, System Sys, Environment Env, ∆𝒕
1 Graph.init(Start)
2 while Graph.size() is less than threshold K
3 Node rand = rand() //Random_State()
4 Node near = Graph.nearest(rand, Graph) // Nearest_neighbour()
5 try
6 Node new = Sys.propagate(near, rand) //New_State
7 Graph.addNode(new)
8 Graph.addEdge(near,new)

29
29
RRT* Search Process
Algorithm RRT
Input : Search space S, limit K, initial state Qnew, goal Qgoal
Begin
tree ← 𝑄𝑖𝑛𝑖𝑡
while goalReached() do
if p < random() then
Qrand ← sampleSpace (S)
Qnear ← findNearest (tree, Qrand, S)
Qnew ← join (Qnear, Qrand, K, S)
Qneargoal ← Qnew
else
Qneargoal ← findNearest (tree, Qgoal, S)
Qnewgoal ← join (Qneargoal, Qgoal, K, S)
addNode (tree, Qneargoal , Qnewgoal )
solution ← traceBack (tree, Qgoal)
return solution
end
4-Jun-21

30
30
MODULE 2 : RAPIDLY EXPLORING RANDOM TREE* (RRT*) PATH PLANNER
DATASET 1
4-Jun-21

31
31
DATASET 1
4-Jun-21

32
32
DATASET 2
4-Jun-21

33
33
DATASET 2
4-Jun-21

34
34
DATASET 3
4-Jun-21

35
35
DATASET 3
4-Jun-21

36
36
DATASET 4
4-Jun-21

37
37
DATASET 4
4-Jun-21

38
HARDWARE IMPLEMENTATION-
OBSTACLE AVOIDANCE

39
39
PHASE 2 – HIGH LEVEL ARCHITECTURE
4-Jun-21

40
40
ROBOT PLATFORM FOR HARDWARE IMPLEMENTATION
NAME X8O SV WiRobot
SENSORS USED 3 Ultrasonic Range Sensors and 7 Infrared Sensors
4-Jun-21

41
41
ROBOT PLATFORM FOR HARDWARE IMPLEMENTATION
4-Jun-21

42
42
SONAR DATA COLLECTION ALONG WITH TIMESTAMP
SAFE ZONE
ALERT ZONE
DANGER ZONE
S1 LEFT SENSOR
S2 MIDDLE SENSOR
S3 RIGHT SENSOR
4-Jun-21

43
43
INFRARED SENSOR DATA COLLECTION ALONG WITH
TIMESTAMP
IR1
FRONT LEFT
SENSOR
IR2
FRONT
MIDDLE
SENSOR
IR3
FRONT
MIDDLE
IR4
FRONT
RIGHT
IR5 RIGHT
IR6 REAR
IR7 LEFT
4-Jun-21

44
44
OBSTACLE DETECTION AND AVOIDANCE

45
45
OBSTACLE DETECTION AND AVOIDANCE
STATIC AND DYNAMIC OBSTACLE
AVOIDANCE
 Move Backwards when the Obstacle is
in front
 Turn Left
 Turn Left to sense the Obstacle and
Moving forward when there is no
obstacle
 Avoid Obstacle and Move Forward
 Avoid Walls and Move Backwards
 Turn Right
 Multiple Obstacle Avoidance
4-Jun-21

47
47
WORKPLAN – PHASE 2
4-Jun-21

49
49
REFERENCES
1. Mohammed Atia, Shifei Liu, Heba Nematallah, Tashfeen B. Karamat and Aboelmagd Noureldin, “Integrated
Indoor Navigation System for Ground Vehicles With Automatic 3-D Alignment and Position
Initialization”, IEEE Transactions on Vehicular Technology, vol. 64, no. 4, pp.1279-1292, April 2015.
2. Marwah Almasri, Khaled Elleithy and Abrar Alajlan, “Sensor Fusion Based Model for Collision Free Mobile
Robot Navigation”, Sensors, vol. 16,no. 24, pp. 1-24, Oct. 2015
3. David Fleer, and Ralf Möller, “Comparing holistic and feature-based visual methods for estimating the
relative pose of mobile robots” , Robotics and Autonomous Systems, Elsevier, vol. 89, pp. 51-74, Dec. 2016.
4. Zhiwen Xian, Xiaofeng He, Junxiang Lian, Xiaoping Hu, Lilian Zhang, “A bionic autonomous navigation
system by using polarization navigation sensor and stereo camera”, Springer Autonomous Robots, pp. 1-
12, July 2016
5. J.Du, C.Mouser and W.Sheng,” Design and Evaluation of a Tele operated Robotic 3-D Mapping System
using an RGB-D Sensor”, IEEE Transactions on Systems, Man and Cybernatics, vol. 46, no. 5, pp. 718-
724,May 2016.
6. S.Lowry and M.Milford,” Supervised and Unsupervised Linear Learning Techniques for Visual Place
Recognition in Changing Environments”, IEEE Transactions on Robotics, vol. 32, no. 3, pp. 600-613, Jun.
2016.
4-Jun-21

50
50
7. R.Yudanto and F.Petre,” Sensor Fusion for Indoor Navigation and Tracking of Automated Guided
Vehicles”, International Conference on Indoor Positioning and Indoor Navigation(IPIN)., Canada, Oct. 13-16,
2016, pp.13-16.
8. Liang Zhang, Peiyi Shen, Guangming Zhu, Wei Wei and Houbing Song,” A Fast Robot Identification and
Mapping Algorithm Based on Kinect Sensor”, Sensors , vol. 15, pp.19937-19967, Aug. 2015.
9. R.Luo and C.Lai,” Multi sensor fusion-based concurrent environment mapping and moving object
detection for intelligent service robotics”, IEEE Transactions on Industrial Electronics, vol.61, no.8,pp.4043-
4051, Aug. 2014.
10. Huizhong Zhou, Danping Zou, Ling Pei, Rendong Ying, Peilin Liu,and Wenxian Yu, ”StructSLAM: Visual
SLAM with building structure lines”, IEEE Transactions on Vehicular Technology, vol. 64, no.4, pp. 1364-
1375, Apr. 2015.
11. Yassen Dobrev, Sergio Flores and Martin Vossiek,”Multi-Modal Sensor Fusion for Indoor Mobile Robot
Pose Estimation”, 2016 IEEE/ION Position, pp.553-556, 2016.
12. F.Zhang, D.Clarke and A.Knoll,” Vehicle detection based on LiDAR and camera fusion”, 17th IEEE
International Conference on Intelligent Transportation Systems, China, pp.1620-1625, 2014.
13. X.Zhang, A.Rad and Y.Wong,” Sensor fusion of monocular cameras and laser rangefinders for line-
based simultaneous localization and mapping (SLAM) tasks in autonomous mobile robots”, Sensors,
vol.12, no.1, pp. 429-452, 2012.
50
REFERENCES
4-Jun-21

51
51
14. O.Wijk and H.Christensen, “Triangulation Based Fusion of Sonar Data with Application in Robot Pose Tracking”, IEEE
Transactions on Robotic Automation, vol. 16, no. 6, pp. 740-752, Dec. 2000.
15. Chunming Yan, Jun Luo, Huayan Pu, Shaorong Xie and Jason Gu, “A Navigation System Based on Vision and Motion
Fusion Information Using Two UFKs”, Proceeding of the 2015 IEEE International Conference on Information and
Automation Lijiang, China, pp. 174-178, 2015.
16. Oscar De Silva, George K. I. Mann, and Raymond G. Gosine, “An Ultrasonic and Vision-Based Relative Positioning Sensor
for Multi-robot Localization”, IEEE Sensors Journal, vol. 15,no.3, pp. 1716-1726, 2016
17. Anthony Spears, Ayanna M. Howard, Michael West and Thomas Collin, “Acoustic Sonar and Video Sensor Fusion for
Landmark Detection in an Under-Ice Environment”, 2015.
18. Chinchu Chandrasenan, Nafeesa T.A, Reshma Rajan, Vijayakumar K, “Multisensor Data Fusion Based Autonomous
Mobile Robot With Manipulator For Target Detection”, International Journal of Research in Engineering and Technology,
vol. 3, no. 1,pp.75-81, Mar. 2014.
19. Lijun Wei, Cindy Cappelle, and Yassine Ruichek, “Camera/Laser/GPS Fusion Method for Vehicle Positioning Under
Extended NIS-Based Sensor Validation”, IEEE Transactions on Instrumentation and Measurement, vol.62, no. 11,
pp.3110-3122, Nov. 2013.
20. Nuno L. Ferreira, Micael S. Couceiro, André Araújo, and Rui P. Rocha, “Multi-sensor fusion and classification with mobile
robots for situation awareness in urban search and rescue using ROS”, IEEE International Symposium on Safety, Security,
and Rescue Robotics, 2013.
51
REFERENCES
4-Jun-21

ROBUST MULTISENSOR FRAMEWORK FOR MOBILE ROBOT NAVIGATION IN GNSS-DENIED ENVIRONMENTS

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to ROBUST MULTISENSOR FRAMEWORK FOR MOBILE ROBOT NAVIGATION IN GNSS-DENIED ENVIRONMENTS

Similar to ROBUST MULTISENSOR FRAMEWORK FOR MOBILE ROBOT NAVIGATION IN GNSS-DENIED ENVIRONMENTS (20)

Recently uploaded

Recently uploaded (20)

ROBUST MULTISENSOR FRAMEWORK FOR MOBILE ROBOT NAVIGATION IN GNSS-DENIED ENVIRONMENTS