Lect 17-19 Flashcards
Reactive Systems in robots
Sense act couples (behaviours)
Behaviour: Relationship between sensorial inputs and motor behaviours to accomplish a task or react to stimulus.
Reactive robots closely links perception to action without any abstract representation or temporal history
What is a reactive paradigm?
sense -> act
In order to have an intelligence you dont need plan
Primitive Input Output
- Sense Sensor data -> sensed info
- We don’t need planning
- ACT Sensed info or directives -> actuator commands
(direct coupling between sense and act)
uses multiple instances of sense act couples which are called as behaviours, that get turned on and off based on stimulus
No plan
IN GENERATING ANSWERS THE USE OF ABSTRACT REPRESENTATION OF KNOWLEDGE IS AVOIDED
What is a stimulus-response
Stimulus -> Behaviour -> Pattern of Motor action
What are the motor actions in 3 categories?
1)Reflexive behaviour (stimulus response) 2)Reactive behaviour (learned) 3)Conscious Behaviour (require deliberative thought)
Examples of behaviours
1)Exploration/directional behaviours (heading bases, wandering) 2) Goal oriented (discrete object attractor,area attractor) 3) Aversive/protective behaviours (avoid stationary objects, elude moving objects) 4) Path following (road following,hallway navigation) 5)Postural behaviours (balance stability) 6) Social cooperative behaviours (sharing,flocking) 7) Teleautonomous behaviours (influence,behavioral modisifaction) 8) Perceptual behavors (visual search) 9) Walking behaviors (gait control) 10) Manipulator spesific behaviours (Reaching) 11) Gripper hand behaviour (grasping)
What is a subsumption architecture?
First reactive architecture. You have multiple behaviour running on parallel with inputs from sensors and output to actuators.
Heavily associated with the behavioral based robotics.
What is embodiment and situatedness?
Embodiment: the intelligence is on the body, the robots body matters.
Situatedness: The environment which the robot is acting is important
Behaviours should take into account both embodiment and situatedness
Structure of behaviour in subsumption architecture
Inputs can be signals or sensors coming into the behavioral module. Inputs can be direct or inhibitive 3 different signals. 1)Input(direct or inhibitor) 2) Reset (if something is wrong reset) 3)Output(can be direct or suppressors)
Why is it called subsumption architecture?
Because there are some layers that are subsuming the nodes of other layers. They are taking over(stopping) the role of other behaviours with the supressor we limit the output of the behaviour.
Behaviours and layered control:
Decision making by set of task accomplisihng behaviours, not done by a deliberative layer.
Each behaviour can be active or non active. Active behaviors can disable the other behaviours.
There is a mechanism for action selection. BEhaviours are organized in layers. Behaviours get active simultaneously.
Because there is no control, AFSMs communicate with each other via inhibition and suppression signals.
Advantage: we can start from lower layers we can test them and as soon as we are happy with them we can add on top another layer
What are concurrent behaviours?
More behaviours are active at the same time and generating output at the same time, so final action is the summation of these actions.
Behaviours are active at the same time, we put them in a sequence, I wond execute them all, sum of them executed before and others executed later.
Subsumption is one of the concurrent behaviours.
Give an example of concurrent behaviour.
Assume we have a robot which is sensing the walls, there are 2 behaviours, avoid left and avoid right wall. The summation of these two will result in go straight.
What is equilibrium?
Behaviours balance each other out
Interaction of concurrent behaviours?
Equilibrium
Dominance of one/winner take all
Cancellation
What are the types of concurrent behaviours?
Competitive:
Only the output of one behaviour is selected
(subsumption,voting)
Cooperative:
blends the outputs of different behaviours to achieve a final goal
(Vector summation and fuzzy logic=
Mixed: A combination of these two (fsa,scripts)