Robotics week6 Flashcards
Sensors and Sensor Fusion, SLAM (20 cards)
What are the four types of sensors used in robotics?
Exteroceptive: Sense environment (e.g. LiDAR, camera)
Proprioceptive: Sense internal state (e.g. wheel encoders, joint angles)
Passive: Receive ambient signals (e.g. microphone)
Active: Emit and detect signals (e.g. sonar, radar)
What key properties are used to characterise a sensor?
Sensitivity
Dynamic range
Resolution
Sampling frequency
Accuracy
Precision
What is the difference between accuracy and precision?
Accuracy: How close a sensor’s readings are to the true value
Precision: How consistent sensor readings are (repeatability)
What is dynamic range, and how is it measured?
Ratio of max to min measurable input
Expressed in decibels (dB)
What is amplitude quantisation and how does it relate to bits?
Signal amplitude is divided into discrete levels
For n bits: 2^n levels
Dynamic range = 20log10 (2^n) dB
ex) 16 bits -> 65536 levels -> 96 dB
What is Nyquist’s theorem and why is it important
a signal must be sampled at least twice its highest frequency
prevents aliasing
What is sensor noise and what are its effects?
Noise from unwanted environmental factors
effects -> can distort measurement
What is the basic idea of sensor fusion?
Combine mulitple sensor readings to improve reliability and accuracy
uses statistic like mean, variance, and probability density functions(pdfs)
In Bayesian inference, which sensor should be trusted more?
The sensor with lower variance(higher precision)
Weight is inversly proportional to variance
What is SLAM problem?
SLAM stands for Simulataneous Localisation And Mapping - it refers to algorithm that track the pose of a robot while building a map of the environment at the same time.
A map is needed for localisation
A localisation estimate is needed to build a map
What are the two main sources of information in SLAM?
- Robot model (prediction) : estimates outcome from actions (using odometry)
- Sensors(update) : refine estimates based on measurement( LiDAR,GPS,IMU..)
Why is sensor data not always reliable on its own
1.Can be noisy or drift
2. Sensors consume power and storage
3. Sensor data may be infrequent
Therefore, we combine sensor and model data using Kalman filter
What is a Kalman Filter?
A reculsive filter that combine noisy model predictions and noisy sensor measurements to estimate a system’s state optimally considering uncertainty.
What does the Kalman Gain represent?
A value between 0 and 1 that determines how much weight is given to the sensor reading versus the model prediction.
High gain (-> 1) : trust the sensor more
Low gain( -> 0 ) : trust the model more
How does SLAM use the Kalman Filter for mapping?
SLAM estimates both robot pose and landmark positions recursively updating both using a Kalman filter as new measurements are received.
What is loop closure in SLAM?
A process where the robot revisits a previously mapped area, detects known landmarks, and reduces accumulated positional uncertainty, improving overall map accuracy.
Name three alternative SLAM algorithms
EKF-SLAM : uses the extended Kalman filter
FAST-SLAM : uses particle filter
Graph-based SLAM: constructs and optimises a pose graph
