Prediction

Question 1

How do we think about handling multi-modal uncertainty?

Answer 1

Maintaining some beliefs about how probable each potential mode is.

Question 2

How are these multi-modal predictions represented?

Answer 2

Represented by a set of possible trajectories such as dotted lines and an associated probability for each trajectory.

Question 3

Name 2 type of Prediction Technologies

Answer 3

1. Model Based

2. Data Driven

Question 4

Explain the Model Based Prediction Technique

Answer 4

Use Mathematical Models of Motion to predict trajectories .

Question 5

Explain the Data Driven Based Prediction Technique

Answer 5

Rely on machine learning and examples to learn from.

Question 6

Is Trajectory Clustering a Model Based or Data-Driven Prediction Technique?

Answer 6

Data Driven Approach

Question 7

Are Process Models a Model Based or Data-Driven Prediction Technique?

Answer 7

Model Based Approach

Question 8

Explain what a process model is?

Answer 8

This is a Model Based Approach.

A process model is a mathematical description of object motion for behavior.

Question 9

Are Multi-Modal Estimators a Model Based or Data-Driven Prediction Technique?

Answer 9

Both.

Question 10

What are Multi-Modal Estimators?

Answer 10

An effective technique for handling the uncertainty associated with prediction, namely, the uncertainty about which maneuver an object will do in a particular situation.

Question 11

Explain the Hybrid Approaches.

Answer 11

Use data and process models to predict motion through a cycle of intent classification where we try to figure out what a driver wants to do. Trajectory Generation tries to figure out how they are likely to do it.

Question 12

A prediction module uses what to generate predictions for what all other dynamic objects in view are likely to do?

Answer 12

A map and data from sensor fusion.

Question 13

Which model is best for determining maximum safe turning speed on a wet road.

Answer 13

In this situation we could use a model based approach to incorporate our knowledge of physics (friction, forces, etc…) to figure out exactly (or almost exactly) when a vehicle would begin to skid on a wet road.

Question 14

Which model is better at predicting the behavior of an unidentified object sitting on the road.

Answer 14

Data Driven. Even with data driven approaches this would still be a very hard problem but since we don’t even know what this object is, a model based approach to prediction would be nearly impossible.

Question 15

Which model is better at predicting the behavior of a vehicle on a two lane highway in light traffic.

Answer 15

Hybrid Approach

Question 16

Trajectory Clustering has two phases.

Answer 16

Offline and Online

Question 17

Define the offline phase for trajectory clustering.

Answer 17

This is where the algorithm learns a model from data.

Question 18

Define the online phase for trajectory clustering.

Answer 18

This is where it uses that model to generate predictions.

Question 19

Name 5 Steps for Target Clustering Offline.

Answer 19

1. Get a lot of trajectories.
2. Clean the data.
3. Define some mathematical measurement of similarity.
4. Perform Unsupervised Clustering.
5. Define prototype trajectories for each cluster

Question 20

Name 3 Steps for Target Clustering Online.

Answer 20

For every update cycle:

1. Observe vehicle’s partial trajectory.
2. Compare to prototype trajectories for each cluster.
3. Predict a trajectory.

1 & 2 comparison is done using the same similarity measurement used for offline clustering.

Question 21

What are Frenet Coordinates?

Answer 21

It is a way representing position on a road in a more intuitive way than the traditional x, y Cartesian Coordinates.

Two main variables: s & d

Question 22

Explain s in Frenet Coordinates.

Answer 22

s is the distance ALONG the road. Also known as the longitudinal displacement.

Question 23

Explain d in Frenet Coordinates.

Answer 23

d represents side to side position on the road. Also known as the lateral displacement.

Question 24

Data-Driven Approaches solve the prediction problem in 2 phases.

Answer 24

1. Offline Training

2. Online Prediction

Question 25

List 3 Steps for Autonomous Multi Modal Algorithm. (AMM)

Answer 25

1. Modal conditioned filtering.
2. Model Probability update.
3. Estimate fusion.

Question 26

With respect to AMM, modal conditioned filtering does what. (Based on the comparative study)

Answer 26

Runs Kalman filter for each model (m) with initial condition.

Question 27

With respect to AMM, the modal probability update does what. (Based on the comparative study)

Answer 27

Evaluates the posterior probability for each model (m).

Question 28

With respect to AMM, the estimate fusion does what. (Based on the comparative study)

Answer 28

Evaluates the overall output - the estimate and the covariance.

Question 29

How many target motion models are used for each multi modal tracking algorithm.

Answer 29

9. 1 non-maneuver model. 8 maneuverable models.

Question 30

Name the 4 discussed in class.

Answer 30

1. Constant Velocity Model - Linear Point Model
2. Non-Linear Point Model - Constant Acceleration w/ Curvature
3. Kinematic Bicycle Model with Controller - PID Controller with distance and angle
4. Dynamic Bicycle Model with Controller - PID Controller on distance and angle

Question 31

What variables make up the AMM algorithm?

Answer 31

1. Consider some set of M process models / behaviors

2. Probabilities for process models (defined as mu)