5a - Robot Navigation

Question 1

What is the definition of navigation?

Answer 1

The process of determining and maintaining a path or trajectory to a goal destination

Question 2

What are 3 key questions in Robot Navigation?

Answer 2

Where am I?
Where do I go?
How do I go there?

Question 3

What does a robot need to be able to be able to answer the 3 questions of Robot Navigation?

Answer 3

A model of the environment

To perceive and understand the environment

To find its position and situation in the environment

To plan and execute the movement

Question 4

What are the 2 ways in which a model of the environment can be provided?

Answer 4

Given or autonomously built

Question 5

Name 2 types of path planning and give an example

Answer 5

Global path planning (graph search) and local path planning (collision avoidance)

Question 6

What is global navigation?

Answer 6

Where the robot is not told its initial position and the position should be estimated from scratch

Question 7

What is position tracking, or continuous localisation?

Answer 7

Where the robot knows its initial position

Question 8

Name 3 types of global navigation strategy

Answer 8

Recognition triggered response, topological route and survey

Question 9

What is recognition triggered response?

Answer 9

Local navigation triggered by the last recognised place

Question 10

What is topological route?

Answer 10

Based on topological maps (graphs where places are nodes and edges are routes connecting them, go geometric information

Question 11

What is Survey?

Answer 11

All known places and spatial relations are embedded in the same frame of reference, the robot can discover new paths

Question 12

Name 4 types of local navigation strategy

Answer 12

Search, direction following, aiming, guidance

Question 13

What is Search?

Answer 13

The robot can move and recognise its arrival at the goal

Question 14

What is Direction following?

Answer 14

Compass direction or trail following (find goal from one direction)

Question 15

What is Aiming?

Answer 15

Taxis to source, using landmarks (can find a salient goal from a catchment area)

Question 16

What is Guidance?

Answer 16

Information by surroundings

Question 17

How can you use vision as a local navigation strategy?

Answer 17

Use a visual cue at the goal reactively, detect goal from a distance and maintain a course towards it, also called landmark navigation/beaconing

Question 18

What are some problems with beaconing?

Answer 18

Cues must be visible, they may be temporarily obscured or have a short range

Question 19

What is the Visual Homing Snapshot model?

Answer 19

Before leaving home take a visual snapshot, when homing move to reduce the mismatch between current views and the initial snapshot, repeat until the home is reached

Question 20

What are 3 problems with the Snapshot Model?

Answer 20

How to define a landmark

How to match landmarks between current scene and memory

You must be aligned

Question 21

What is loop closure?

Answer 21

The ability to recognise a location even if perceived from a different pose (also known as the kidnapped robot problem)

Question 22

What is feature based place recognition?

Answer 22

Using visual features to describe a location

Matching the features and descriptors current visible with a list of those at previously visited locations

Also called bag of words

Question 23

What is the ALV visual homing model?

Answer 23

Average Landmark Vector

Before leaving home take a snapshot, point a unit vector to each surrounding landmark, compute ALV and store

Then use vector subtraction to generate the home vector

home vector = current ALV - home ALV

Question 24

Name 3 possible navigation strategies for robots

Answer 24

Navigation by vision and compass

Navigation using landmarks as beacons

Navigation by position tracking

Question 25

Discuss the requirements of a general-purpose solution to the navigation problem in mobile robotics

Answer 25

A representation of the navigable space (as a graph or grid derived from global map of environment)

A path planner (A*/Dijkstra algorithms) to find best route to goal location

An autopilot to drive between waypoints taking into account the robot’s kinematics and dynamics

Avoid unexpected obstacles on the way