Sensor Fusion

1
R2D2 - SENSOR FUSION
SENSOR FUSION GROUP 1 SOFTWARE ENGINEERING
FINAL PRESENTATION
2
Members
R2D2 are Jesper Simos Jesper Sφderlund Quang
Tien Le Ricky Stanley DCruze Sreedhar
Danturthi Thomas Sφrensen Vasilis Odontidis
3
Agenda
1. Project Results 2. Project Experience 3.
Requirements and design
4
1. Project Results
5

• Total hours of work 1200
• Most hard working guy Jesper Sod.

6
Project Cost
In money Actual Cost 600,000 Kr Planned Cost
700,000 Kr
7
Project Status
Activity W14 W15 W16 W17 W18 W19 W20 W21 W22
Project Description
Requirement
Prototype
Design Description
Implementation
Testing
Final Product
MIL!
MIL!
MIL!
MIL!
MIL!
MIL!
MIL!
8
R2D2 Deliverables
Name Promised Delivered
Project Description 15 16
Requirements 16 17
Prototype 17 18
Design Description 18 22
Test Specification 22 22
Final Project Report 22 22
Final Product 22 22
9
2. Project Experience
10
Project Experience 1

• The project evolved quite similar to the project
  plan, apart from some deviations
• in milestones.
• More time on design than expected due to
  constant and extreme changes which
• affected implementation.
• Less time on actual implementation.
• In the beginning synchronization and
  distribution of work was very smooth according
• to initial requirements.

11
Project Experience 2

• ADA 95
• Robots playing soccer
• Robotics in General
• Team Work
• Communication
• Respect other peoples ideas
• Others might know better than us

12
Problems in retrospect

• Deliverables and communication with Robolab.
• Unavailable customer in critical time.

13
REQUIREMENTSAND SYSTEM DESIGN
14
Functional Requirements

• Identify Robot position
• Identify Robot speed
• Identify Team members Position and Speed
• Identify Ball Position
• Identify the balls speed and direction
• Identify Opponent robots position
• World View
• Kalman Filter implementation
• Sensor Fusion
• Testing

15
Non-Functional Requirements

• Cost Goal
• Time Goal
• Data provided in real-time with high accuracy
• Operating system, programming language and
  software features

16
Use cases
17
Overall system design
18
Kalman Filter
The Kalman filter is an efficient recursive
filter that estimates the state of a dynamic
system from a series of incomplete and noisy
measurements.
Wikipedia Kalman Filter
19
Discrete Kalman ?lter time update equations X
Ax Bu - P APTranspose(A) Q
Discrete Kalman ?lter measurement update
equations K P Transpose(H) Inverse( H
P Transpose(H) R ) X X K ( Z H
X ) P ( I K H ) P
20
Example Filtered Values (x,y,Rot) Should be
x, y 10.00 Rot 90 14.43, 11.25,
91.21 (x,y,Rot) Should be x, y 20.00 Rot
90 21.11, 25.58, 93.26 (x,y,Rot) Should be
x, y 30.00 Rot 90 35.33, 33.00,
93.54 (x,y,Rot) Should be x, y 40.00 Rot
90 44.63, 42.23, 94.71 (x,y,Rot) Should be
x, y 50.00 Rot 90 53.84, 53.55,
91.54 Actual Measurements History (x,y,Rot,Rot
,x,y,Rot) 25.92, 12.01, 90.24, 98.49,
13.06, 11.16, 90.89 (x,y,Rot,Rot,x,y,Rot)
21.01, 38.45, 90.31, 92.05, 21.13, 24.04,
93.68 (x,y,Rot,Rot,x,y,Rot) 46.06, 31.28,
91.61, 92.29, 34.05, 33.21,
93.84 (x,y,Rot,Rot,x,y,Rot) 49.64, 49.74,
97.65, 96.86, 44.03, 41.34,
94.23 (x,y,Rot,Rot,x,y,Rot) 66.93, 68.79,
95.46, 91.82, 52.27, 51.73, 91.05
21
4. Demonstration
TO BE CONTINUED
22
THANK YOU!!!