1. Problem
   1. What driving style should an auto car have?
   2. Specific driving behavior for each driving situation
2. Goal
   1. Learning driving style preference while satisfying safety constraints
   2. From the study
      1. Teach auto car how to drive from human demonstrations
      2. How do driving styles relate to comfort
3. Opportunity
   1. Achieve safe driving with auto car and satisfy driver's style
4. Significance
   1. Impact on auto car driving behavior
   2. Increase driver satisfacation
5. Related Work (Existing Studies)
   1. General driving style
      1. Driving speed
      2. Tendency of committing traffic violation
      3. Headway
      4. Overtaking of other vehicles
   2. Most common metric to define style
      1. Aggressiveness
      2. Defensiveness
   3. Multidimensional driving style inventory (MDSI)
      1. Reckless and careless driving----High speed
      2. Anxious driving
      3. Angry and hostile driving (More use on horn and flash)
      4. Patient and careful driving
   4. Huysduynen
      1. Angry driving
      2. Anxious driving
      3. Dissociative driving
      4. Distress-Reduction driving
      5. Careful driving
   5. Lee (Analyzed lane changed)
      1. Severity
      2. Urgency
      3. Type classification
   6. Hong
      1. Differentiated styles in term of defensiveness
      2. Propensity for rules violation
   7. Horswill
      1. Valuable distinction between skill and style
   8. Scherer
      1. Comfort
6. Assumption
   1. Focus on defensiveness of driving style
   2. An auto car should learn their user's driving style/ behavior
   3. Driving defensiveness
      1. Aggregate of driving features in various driving scenarios
7. Study
   1. Part 1
      1. Survey of participants driving behaviors
      2. Test track info gatherings
         1. 9.6 miles track
         2. 15-20 mins drivings
   2. Part 2
      1. 6 Tests
         1. Lead car slow down forcing lane change
         2. Merge back to right lane
         3. Slow lead car in right lane forcing lane change when another car is approaching fast in the destination lane
         4. Merge back into right lane with a continuous traffic moving at constant gap and constant speed
         5. Left turn at stoplight and yield at green
         6. Right turn on green light
      2. 4 driving styles simulated
         1. Aggressive
         2. Defensive
         3. Own drivings
         4. Another participants drivings
      3. 9 out 15 preferred different style than their own at least one of the tasks
      4. 46%-67% could not identify tasks corresponds to their behavior
      5. Inner city, highway simulation
      6. Randomized tasks
   3. 15 participants (college students)
   4. 46% considered themselves well experience driving, and 20% considered themselves experienced
   5. Mean driving experience was 5.46 years
   6. Rest 18-24
   7. 3 participants were 30 to 31 years old
8. Limitation
   1. Limited driving style features
   2. Limited driving style choices
   3. Limited fidelity of simulation environment
9. Evaluation (Result)
   1. Simulation
      1. Similarity between driving on road and driving in our study simulator was +1 on -3 to +3 scale.
         1. Simulator conditions are very similar to real driving conditions
   2. Feature distribution for participant styles
      1. Define driving style in terms of features, for each task, feature, and participant, what the participant's feature value was for that task.
         1. Higher negative values correspond more aggressive behavior
   3. Preferred style in relation to own style
      1. Future research
         1. driving style based on context
   4. Perceived own style in relation to actual own style
   5. Should test more diverse feature choices and driving style representation in a higher fidelity setting
   6. Explore what features users' consider when they evaluate autonomous driving styles
   7. Explore new learning technique that can augment user demonstrations with other types of user input and guidance