7. Increasing Demand of Robots
• Robot will be required to operate in homes and
in hospitals to service the rapidly aging population.
• Will be required to mine and farm in increasingly
remote areas.
• Robots will need to perform tasks beyond their
explicitly pre-programmed behaviors and quickly
adapt to variable nature of tasks
8. To perform all these tasks Robot will need to carry powerful computers
and large batteries to power them .
10. Navigating to the Grasping and picking
Locating the bottle up the bottle
bottle’s position
Robot A
Giving the bottle to Navigating towards the Locating the patient
the patient patient
11. Queries the
Downloads the
Robot B database for
knowledge
relevant information
Sends it’s experience Perform the specific
to the database task
12. • The information downloaded provides a useful starting point.
• The detailed object models can increase the speed and
reliability of Robot B
• A set of collected information can guide the 2nd robot’s search
and execution strategy.
• As the two robots continue to perform their tasks and pool
their data, the quality of prior information will improve and
begin to reveal underlying patterns and correlations about
robots and their environment.
• The availability of such prior information is a necessary
condition for robots to operate in a more complex,
unstructured environments.
16. Server
• Core layer
• Holds database
• Accessible via common web interfaces
• Contains CAD models, point-clouds and image data
• Maps are saved as compressed archives
• Robot task description stored as human readable action recipe
• Action recipe arranged in a hierarchy
17. Server
e
Ob e cip
n r ns
•T jec on
o
cti trai
•D ype
•L ime s ts cti u
a
b- ons ter
f s g c me
oc n
at sio
ion n s
Environment A o
ist erin par
a
•Maps for • L rd n
• O ctio
self-localization •A
18. Interfaces
1. Web interface
2. Based on Representation State Transfer
(RSET) architecture.
3. Based on publish/subscribe message
exchange pattern
19. Generic Components
Action Execution :
Determining and checking the availability of a set of skills required for the task.
Finding an ordering among them that satisfies the ordering constraints.
Linking a robot’s perceptions to the abstract task description given by action recipe.
Reliable action execution.
20. Environment Modeling
1. World model component combine prior information available via database
with robot sensory input
2. To guarantee fast and safe performance it uses a two-level approach.
Semantic Mapping
• RoboEarth database provides generic action recipes defined in terms of
robot’s skills, its environment, and the objects it is going to interact.
21. Learning
1. Repeated use of action recipes an skills offers many opportunities to improve the
reliability and performance of action execution.
22. Robot Specific
Action and situation recognition and labeling
• The generation of useful, transferable action recipes and their
successful reuse is challenging task
• The recognition and labeling component is designed to simplify
the generation of new action by abstracting from the robot-
specific low level layer and aggregating subsequent steps that
belong to one complex task into an abstract task description.
• Essential for learning new recipes by demonstration.
23. Skill Abstraction Layer and Robot Specific Components :
• Provides a generic interface to robot’s specific, hardware-
dependent functionalities.
24. Benefits
• Provides a shared knowledge i.e. organizes and unifies
information about the world
in a format usable by robots.
• Offloads heavy computing tasks to the cloud.
• Longer battery life.
• Less need for software upgrades are invisible and hassle-
free.
• Reusable library of “skills” or behaviors that map to
perceived task requirements / complex situations.
• Data-mining the history of all cloud-enabled robots.
25. I need a helicopter pilot program….
Trinity from Matrix