Robotics week3 Flashcards
Control-Theory, PID-Control (29 cards)
What is the main purpose of a regulatory/control system?
To maintain a desired state (intention) by sensing the current state, comparing it to the goal, and acting via feedback to reduce any discrepancy—often referred to as homeostasis or closed-loop feedback control.
What characterizes an open‐loop control system?
In open‐loop control, the agent issues actions based on intentions without using sensory feedback to correct errors. It relies entirely on a predefined transform or inverse model(로봇), so disturbances or changes in the world are not compensated for automatically
How does open‐loop control using an inverse model(robot) differ from dead‐reckoning?
Dead‐Reckoning (no model): Actions are issued via trial‐and‐error mapping from intentions to actions, but any disturbance causes drift.
Inverse Model (robot)(inverse kinematics): Uses a computed inverse of the forward transform (fw⁻¹) to generate actions expected to achieve intentions, but still cannot correct for unmodeled disturbances
What defines a closed‐loop (negative‐feedback) control system?
Continiously senses the outcome of actions, compares observed consequences to intended goals, and adjusts actions to minimize the error. It automatically compensates for disturbances in the world model (fw) or environment.
Describe the step‐by‐step process of closed‐loop control (feedback).
1.Set the intended result (reference input).
2.Sense the current situation (feedback from sensor).
- Compare sensed value to the intended value, computing the error.
4.The controller issues actions to reduce the error.
5.Repeat the loop until consequence equals intention.
Describe the step-by-step process of open-loop control(dead reckoning)
- Issue command
- action initiated
- effect may or may not be as intended
what is servomechanism and how does it use feedback?
A servo is an actuator coupled with a feedback sensor and a controller.
The reference input specifies the desired position or speed, the sensor measures actual state, the controller adjusts the actuator commands to minimize the difference, achieving precise control of motion
Why is negative feedback important for stability?
It continiously counteracts disturbances
What is Perceptual Control Theory(PCT)
PCT emphasizes that behaviour is ultimately directed towards maintaining a particular perception, rather than a pre-determined action or outcome
How does hierachical feedback control in PCT differ from simple subsumption?
Subsumption:
Layers of reactive behaviors run in parallel/
higher layers suppress lower ones but do not feed hierarchi perceptual references
Hierachical PCT:
Each layer is a closed-loop controlling a percept.
Higher layers set goals for lower layers, forming a nested architecture of control rather than mere suppression.
List the first four hierarchical levels in Perceptual Control Theory
1st‐order (Intensity): Control of raw signal intensity (light, loudness).
2nd‐order (Sensation): Control of perceived qualities (brightness, pitch)
3rd-order (configuration):
Control of spatial or structural relationships(shape)
4th order(Transitions): Control of changes or movements between configurations
Give one advantage of closed-loop control over open-loop control
Automatically corrects for disturbances and module inaccuracies, maintaining the desired output even when the environment or system dynamics change.
What’s the pros and cons of layered control(PCT) compared to subsumption?
Pros:
1. Parallel execution with closed-loop stability at each level.
2. Better disturbance rejection due to nested feedback
Cons:
1. Difficult design beyond two levels
2. Lack of widely accepted design principles makes hierachy engineering challenging
What role do sensors play in closed-loop systems
Sensors meausre the system’s actual output (the “consequence”) and feed this information back to the controller, enabling it to compute the error and adjust actions.
in PID controller,
How do the three gains Kp, Ki, and Kd individually influence system behaviour?
Kp (Proportional) : Increases reduce rise time but can amplify overshoot or cause oscillation if too large.
Ki (Integral) : Eliminates steady state error by integrating error over time, excessive Ki causes oscillations or “integral wind up”
Kd (Derivative) : Dampens overshoot and improves stability by responding to error rate -> sensitive to sensor noise
What is a key design challenge when stacking control loops hierarchically?
Tuning gains and time constants across multiple nested loops without causing instability or unwanted interactions is difficult; if higher-level loops react too aggressively, they can destabilize lower-level loops, leading to oscillations.
What happens with a P-only controller, and how does varying Kp affect performance?
P-only yields u(t) = Kpe(t). It cannot eliminate steady- state error -there will always be a residual offset.
if Kp = 0, no corrective action
small Kp : Slow response & large error
large Kp : faster response but large possible oscillation
What are the primary limitations of PID control
- Fixed Parameters: Gains Kp, Ki, Kd cannot adapt automatically to changing dynamics
- No plant Model usage: Treats the system as a “Black box”, cannot leverage known process structure
- Linear, Symmetric Response: Struggles with strongly non-linear or asymmetric systems
- Derivative Noise Sensitivity : The D term amplifies sensor noise
- Integral windup (integral term continues to increase even when the controller output is saturated)
: If the actuator(작동기 -> 시스템을 움직이거나 제어하는데 쓰이는 기계 장치) saturates(reaches the maximum limit) the integral term keeps accumulating error - Single-Layer Complexity: One PID can’t handle multi-mode tasks without additional adapation.
