SSS: A Hybrid Architecture Applied to Robot Navigation

1
SSS A Hybrid Architecture Applied to Robot
Navigation
Review Paper By Kai Xu
Whats this?

• Jonathan H. Connell
• IBM T.J. Watson Research Center

2

• SSS Servo, Subsumption, Symbolic
• Servo Control component
• Subsumption Deal with Multi-task
• Symbolic complicated task, like routing/building
  internal model

3
Background

• conventional servo-systems(SPA)
• Inner model is not accurate enough
• Only good for simple tasks
• multi-agent reactive controllers
• i.e. subsumption, behavior-based system
• Fast response, but slow sample rate, discrete
  space
• state-based symbolic AI systems

? Can we combine them?
4
Architecture of SSS
5
Components

• Servo Layer
• continuous time and continuous space
• Behavior-based Layer
• Has special purpose recognizers
• Response to certain situation
• Symbolic Layer
• Triggered by certain events and then response

6
Interface

• Servo and Behavior-based Layer
• Command interface
• setpoints
• Sensor Interface
• Matched filter
• Behavior-based and Symbolic Layer
• Command interface
• Turn on/off, Parameterize modules
• Sensor Interface
• Event triggering
• Contingency table

7
Task of Robot

• Map the environment
• Navigate itself during moving(self-adaptibility)
• compensating for the variability of the
  environment
• Solution use coarse map
• Avoid loop knowing when the robot has arrived
  back in a place it has been before
• Solution memorize the location

8
Tactical Navigation

• Implemented in ServoSubsumption layer
• servo layer
• consists of two velocity controllers, one
  for translation and one for rotation
• Wall-following
• Subsumption layer
• two tactical navigation modules based on
  odometry.
• Slows or stops the robot when approaching the
  destination
• Control the average heading of the robot

9
Strategic Navigation

• Implemented in Symbolic Layer
• Maintaining a coarse map
• Landmarks for IR to detect in long distance
• Paths and their relationlength
• For routing, f.g. spreading activation algorithm
• Not happened very often, only when some event
  happen, f.g. reach an intersection

10
An Item in Contingency table

• (do-until-return
• (setq recognizers (check-situations))
• (cond (
• (and (near-distance? recognizers)
• (aligned-okay? recognizers)
• (left-opening? recognizers)
• )
• (inc-heading! 90)
• (new-travel! 564)
• (return recognizers))
• ((beyond-distance? recognizers)
• (inc-heading! 180)
• (new-travel! 48)
• (return recognizers))
• ((no-progress? recognizers)
• (disable! stay-aligned)
• (enable! scan-for-escape)
• (return recognizers))
• (t nil)))

11
Experiment

• What might happen to the sensor

So What to do?
12
Experimental results
What is felt by robot
What is built in mind
13
Solution

• Using landmarks
• Locate itself dynamically
• Avoid errors accumulation
• Reduce the requirement of accuracy and stability
  of the lower layers more flexible
• Dont need precise time or speed control
• Dont need precise motor or steer control

14
Comparison

• ATLANTIS AuRA
• Similarities
• three layers with increasing deliberative
  abilities
• Interface between layers
• Differences
• No global objective in subsumption layer in SSS
• No internal state in subsumption layer in SSS
• Symbolic still paticipate the control system
  directly, but only for those important events

15
Conclusion

• Advantages
• Sensor the world dynamically, more accurate
• Make use of landmark, more stable
• Finish the task by a coarse map, more practical
• Problems
• Highly rely on Landmarks
• Need prior knowledge

16
Thanks

• Question?