M03 - Controllers (Brains) in Autonomous Systems Flashcards
What is situatedness?
- existing in a complex, dynamic and unstructured environment that strongly affects the robot’s behavior
Explain the loop that generates behavior of robots.
- the robot determines the state of its actuators on the basis of the current and previous observations
- the action of the robot alters the robot/environmental relation and the environment itself
What is the sensory-motor coordination?
The ability of an agent to act so to ensure that it will later experience useful observation
i.e. sensory states enabling the robot to achieve its goals
What are the steps of a sensory-motor coordination?
actions performed by robot -> effects further observations -> determine next actions
Can you predict a dynamic system, even with a complete knowledge of the robot & environment?
No (that is why they are hard to design)
Why do we need controllers?
- real world is dynamic, unpredictable
- to make a decision to accomplish a task, the situated robots need a controller
- make decisions based on perceived sensory information to decide
What control models do we have?
- open loop control
- closed loop control
Explain the open loop control model.
- parameters of the algorithm are set in advance
- parameters do not change while the system runs
- the desired results are (often) not guaranteed
Explain the closed loop control model.
- data from sensors influence behavior
- system uses errors to generate actions (goal is to minimize error for desired state)
What control model do autonomous robots use?
closed loop control
Explain the open loop control through an example
e.g. microwave
- semi-autonomous
-needs user input
- if the timing is not correct given by user, you can burn meal
- the system does not know the desired output to adjust timing by itself
What two complementary kinematics do we have?
- forward kinematics
- inverse kinematics
Explain the forward kinematic.
Given a sequence of commands, what is the final position of the robot’s arms?
- easy to compute
Explain inverse kinematics.
Given a desired position of the robotic arm, what sequence of commands will bring it to that position
- hard to compute
- there can be one, multiple or no solution
What components are there in the closed-loop control system aka feedback control?
r: reference value
y: output
u: control value
e: error
Explain the reference value r in a feedback loop and give an example.
The specification of robot’s task.
e.g. the position of robot relative to a stack of shelves and the distance of gripper
Explain the output y in a feedback loop and give an example.
It is the actual state of the robot
e.g. the distance to an object
Explain the control value u in a feedback loop and give an example.
When the output is fed back to the control algorithm then you can compute control value
e.g power setting of the motor for a robot to get to a shelf
Explain the error in a feedback loop.
When the output is compared to the reference value, then you can compute the error
How does the On-Off control work?
we need constant reference value e.g. distance to an object at which robots need to stop
What control algorithms do we use?
- closed-loop control system
- on-off control
- proportional control
- proportional-integral controller
- proportional-integral-derivative controller
What is the approach of a proportional control system?
“Reduce speed more as the error between reference distance and measured distance gets smaller.”
What is gain in proportional control?
The factor of proportionality of the control algorithm
What are the types of gain?
- high gain
- low gain
- excessive gain
What is high gain?
The robot approaches the reference distance faster
What is low gain?
The robot approaches the reference distance slower
What is excessive gain?
The controller behaves like bang-bang algorithm aka unstable controller
What is a paradigm?
A basic framework of assumptions, principals and methods from which the members of the community work.
What world assumptions do we have?
- open world assumption
- closed world assumption
Describe the closed world assumption.
The robot has an access to world model that contains everything that robot need to know to accomplish the task.
Describe the open world assumption.
The world model is dynamic the robot can not access everything that robot need to know to accomplish the task.
Describe the deliberative architecture.
Agent architecture to be one that contains an explicitly represented, symbolic model of the world and in which decision are made via logical reasoning, based on pattern matching and symbolic reasoning.
What is the reactive paradigm?
- Connects actions with sensing
- Sensors produce continuous signals that, after some manipulation can drive the robot’s motor behaviour to act
Do Braitenberg vehicles follow a reactive paradigm?
Yes.
What are the advantages of the reactive paradigm?
- it is fast
What are the disadvantages of the reactive paradigm?
- not being able to save information
- not being able to form internal representations
- not being able to learn
What does it mean to act deliberately?
performing actions that are motivated by some intended objectives and that are justifiable by sound reasoning.
Do autonomous robots need deliberation in a fixed, well-modeled environments?
No
What are the different types of deliberation functions?
- planning
- sensing
- acting
- monitoring
- learning
- reasoning
What are the component of a hierarchical (deliberative) paradigm?
- Sense (provide a representation of environment)
- Plan (logic/neural/rule-based engine to decide actions)
- Act (the robot’s motor behavior)
What are the advantages of a hierarchical (deliberative) paradigm?
- accurate
- act strategically
What are the disadvantages of the hierarchical (deliberative) paradigm?
- slow, needs sufficient time to generate a plan
- dynamic characteristic of the environment
Describe the hybrid deliberative/ reactive paradigm.
Combines hierarchical/deliberate and reactive paradigms
What are the steps of the hybrid deliberative/reactive paradigm?
- Hierarchical/deliberate -> Plan
- Reactive -> Sense-Act
