Robotics week4 Flashcards
Reaching and Grasping / Local-Guidance (17 cards)
Whas is the primary task in kinematic control of a robot arm?
The goal is to control the end-effector’s(hands) position by coordinating the robot’s joint angles and linkages. Forward kinematics computes end-effector pose from known joint angles; inverse kinematics computes joint angles needed to reach a desired end-effector pose
Why is kinematic control of a robot arm considered “open loop”?
Inverse kinematics generates joint-angle commands from a target position without directly sensing the end-effector(손)’s actual pose.
List three(3) sources of error in open-loop kinematic control
Sensor Errors: Encoders can drift, fail, or be unable to directly measure end-effector pose
Positional Errors : The actual end-effector position may not match the target due to mechanical tolerances and blacklash
Controller/Model Errors: System performance degrades over time, and kinematic models are never perfectly accurate/
What is visual servoing, and how does it overcome limitations of open-loop kinematics?
Visual servoing adds a camera to the robot to observe the target directly. By comparing the current camera view to a stored target view, the system computes image-space errors and translates them into joint-space cocrective movements thus closing the loop around end-effector position.
What is the image Jacobian?
It allows one to compute how the image feature moves in pixel space given a known camera motion
What are the main advantages of using soft grippers and haptic sensing(sense o in robot grasping (잡는것)?
Soft Grippers: Their compliance allows the robot to conform to varied object shapes without precise planning, reducing the need for exact pose estimation .
Haptic Sensors: By detecting light touch, stretch, or pressure, they provide feedback when grasping delicate or occluded objects.
What are the three main search strategies used when no guidance cues are available?
- Brownian walk: continuous motion with small, random turns
- Levy flights: step sizes drawn from a long- tailed distribution, producing clusters of local searches interspersed with long jumps.
- Spiral search: systematic, expanding sprial pattern to exhaustively cover an area.
Describe Brownian (random) walk and its typical applications
In a Brownian walk, the agent moves in roughly straight lines for short distances, then make random turns.
Common in both particle diffusion and some animal foraging, but it may not be optimally efficient for sparse target distributions.
What characterizes a levy flight, and why can it outperform Brownian motion?
Levy flights use a heavy-tailed distribution for step lengths, so most moves are short but occasional long jumps occur
This pattern creates clusters of intensive local searching interspersed with rapid relocations.
Explain spiral search and how hybrid strategies improve coverage.
Spiral Search has the robot move outward in a tight, expanding spiral when it has no prior map - ensuring exhaustive coverage.
Hybrid approaches combine local spiral loops with edge-following or occasional long jumps to reduce redundant coverage and handle obstacles more effectively.
What is beaconing in the context of local guidance?
Beaconing is directed movement toward a detectable cue or beacon
The robot continuously senses cue intensity or direction and adjusts its trajectory to move up the “gradient” of the signal, ultimately homing on the source.
How do Braitenberg vehicles illustrate reactive beaconing behaviour?
Braitenberg vehicles wire sensors directly to motors with excitatory or inhibitory(억제적인) connections.
By adjusting excitatory/inhibitory and crossed/uncrossed wiring, different taxis behaviours (attraction, repulsion, aggression, cowardice, love) emerge.
What is visual homing, and how does it close the loop around end-point location?
Visual homing stores a “snap shot” image at the goal location
On return, the robot captures its current panoramic view and computes a control signal that minimise the image difference between the current view and the stored snapshot driving it back toward the goal.
What key assumptions underlie visual homing approaches?
- Unique feature positions : Each visual landmark appears at a specific location in the visual field only when the robot is at a particular spot.
- Predictable feature motion: As the robot moves, changes in landmark positions within the image follow a predictable pattern
These assumptions allow the robot the select movement directions that decrease the difference between its current view and the snapshot
What are Image Difference Function(IDFs), and why are they useful in visual homing?
IDFs compute a pixel-wise difference between the current panoramic image and the stored home image.
As the robot moves, the RMS difference increases or decreases smoothly, provideing a gradient it can follow “downhill” to minimize the difference.
Advantages: 1. No explicit feature matching is required
2. Good for complex environment.
3. feature-rich scenes where geometric correspondence is hard.
When is IDF useful?
- Long-range line of sight provides long distance guidance
- Short line of sight requires many memories
- Final approach to specific location (automated piloting)
How to find a home vector with Average landmark vector?
Home vector = ALVnow - ALV home
