Mobile robot mapping and navigation

1
Mobile robot mapping and navigation

• Marin Kobilarov
• Trinity College, Computer Science
• Senior Seminar
• Spring 2003

2
Mobile Robots

• Humanoid Robots

wheels six legs
snake
3
Applications
Excavation and Rescue
Personal Assistants
Many more
Space Exploration
4
Mapping and Navigation

• Mapping
• knowledge of environment
• knowledge of position
• Navigation
• path planning
• collision avoidance
• fast re-planning
• Mapping Navigation ? Autonomous Mobile
  Platforms

5
Goals

• Develop mapping and navigation system
• Indoor office like environment
• Wheels
• Proximity sensors
• Existing State-of-the-art algorithms
• New hybrid mapping algorithm
• Integrate mapping and navigation on a real robot

6
Project Challenges

• Theoretical
• Complex probabilistic models
• Machine learning algorithms
• Probability, Numerical Analysis, Linear Algebra
• Software
• Scientific programming
• Design and implementation of simulation
  environment
• Practical
• Transfer system to an actual mobile robot

7
Mapping

• map

• Uncertainty in sensor data
• Uncertainty in odometry
• Uncertainty in environment

8
Mapping Problems

• Representation
• Landmarks
• Grid
• Point obstacles
• Convergence
• Incremental (real-time)
• Requires position
• Map cycles
• Correspondence
• Dynamic environment (beyond scope)

9
Existing Approaches

• Kalman filter
• No correspondence
• Expectation Minimization
• Not incremental
• Hybrid
• No convergence
• Occupancy Grids
• Requires knowledge of position
• Others
• worse

10
Hybrid Approach

• Hybrid w/ Markov Localization
• Hybrid with extended local EM alignment
• Real time
• Maps cycles
• HELEMA w/ Particle Filters
• Particle filters

11
Path Planning

• Value Iteration
• Steepest descent search
• Modified Value Iteration
• Multiple goals
• Modified D Value Iteration
• Multiple goals
• Occupancy Grid
• Potential Fields for obstacle clearance

12
Exploration

• Nearest unexplored
• naive
• Greedy exploration
• Entropy filter
• Greedy Nearest unexplored w/ path replanning
• optimal

13
Robot Implementation

• System models ? Robot specs
• Dynamics
• Odometry
• Sensors
• Wireless link
• transparent half-duplex radio transceiver 33.6
  kbps
• link layer protocol with ISO 3309 error
  correction
• real-time asynchronous application level protocol