September 28th, 2004

1. CICAS Coordination Meeting Virginia Update September 28th, 2004

2. Research Questions Addressed To Date • Is there a difference in brake profiles for distracted vs. willful vs. baseline drivers? • What’s a too early warning? • What’s a too late warning? • What timing aspects of the algorithm will minimize false alarms and misses? • What is the effect of warning mode on driver response? • How many detection points are needed to have an effective algorithm? • What functional requirements are known?

3. Is there a difference in brake profiles for distracted vs. willful vs. baseline drivers?

4. Braking Profiles • Algorithms can be built based on detection of violation-likely groups: • Drivers stop differently depending on their intentions and level of distraction • Distracted drivers stop harder than others • Distracted drivers are less likely to stop • Distracted drivers are more likely to violate • Willful drivers tend to speed

5. What’s a too early warning? • A warning that is issued to a driver that would have stopped without any intervention • Creates annoyance • Decreases user trust • Too early was determined during first three intersection studies

6. Too Early Distribution • The figure depicts the distance from the intersection at which baseline drivers initiated braking when the signal change occurs at 185’ • An algorithm that initiates a warning prior to reaching 135’ would create false alarms Number of Drivers (out of 28) Distance to Intersection (ft)

7. What’s a too late warning? • A warning that is issued to a driver that would benefit, but with timing such that insufficient distance remains for the driver to perceive, react, and stop prior to entering the intersection. • Decreases safety benefit of system • Decreases user trust

8. Ideal Case – Classification Clearance

9. Real Case – Classification Interference

10. Maximizing Curve Separation(Methods to Minimize Misses and False Alarms) • Take advantage of the “all red” phase and the time it takes for opposing vehicles to get into the collision zone to allow vehicles to pass through without warning (time) • Take advantage of the intersection’s buffer zone, the area beyond the stop bar but prior to significant collision risk (space) • Design warnings that minimize reaction time and maximize deceleration by conveying necessary urgency

11. “Normal” Brake Initiation Normal ApproachNo Warning or Timely Warning

12. Algorithm Trip “Normal” Brake Initiation Reaction Time Too Late Warning

13. Allowing the ViolationPreventing the Crash

14. Making Use of Pre-Collision Zones of the Intersection Collision Zone Intrusion Zone Violation Zone Compliant Zone

15. Warning Activation Preventing the Collision

16. Maximize Warning Effectiveness • DII/DVI research to date has demonstrated the importance of countermeasure design. • Prototype warnings were evaluated for Urgency, Distinguishability, and Appropriateness in the lab • Evaluators preferred icon auditory warnings over descriptive (i.e. buzzer vs. “Stop”) • However, experimentation showed clear advantages for the word “Stop”

17. Maximize Warning Effectiveness

18. Comparison of Auditory Alerts

19. Effectiveness of Visual Display • High Heads-Down Visual DVI was ineffective • The display was perceived by less than 5 percent of the participants • Evaluation of DVIs is now focused on auditory and haptic warning modes

20. How many detection points are needed to have an effective algorithm? • Single-Point detection of speed does not result in reliable warning decisions • Improved reliability would consistently result in too late warnings • Continuous detection is the most adaptive to any algorithm type and produces the best theoretical performance • Multi-point alternatives are being tested

21. Simulation of Single Point vs. Continuous DetectionMissed Violations • Speed determined at given distance from intersection for violations that occurred • Single point detection would not have worked for most of the drivers who violated • With continuous detection, three violating drivers would not have been detected; however, none of these drivers would have been in the crash zone.

22. What functional requirements are known? • Detuning tests run to date • Sensors • Acceleration • Velocity • Positioning • Lateral Position • Longitudinal Position

23. Sensor: VelocityNormalized Deceleration: At DVI onset, the average deceleration in g’s that would be required to stop by the stop bar

24. Sensor: VelocityNormalized Deceleration: At DVI onset, the average deceleration in g’s that would be required to stop by the stop bar

25. Sensor: AccelerometerNormalized Deceleration: At DVI onset, the average deceleration in g’s that would be required to stop by the stop bar

26. Positioning: LateralNormalized Deceleration: At DVI onset, the average deceleration in g’s that would be required to stop by the stop bar

27. d = 125.6 ft D = 125 ft Positioning: Lateral • Assumes a 12 foot lane width • Would also apply to curved road geometry

28. Positioning: LateralLane Position: Correct vs. Incorrect

29. Positioning: LongitudinalNormalized Deceleration: At DVI onset, the average deceleration in g’s that would be required to stop by the stop bar

30. Positioning: LongitudinalNormalized Deceleration: At DVI onset, the average deceleration in g’s that would be required to stop by the stop bar

31. Project Plans • Completing studies to answer the following questions • What is the optimal timing for collision warning? • What is the optimized braking profile for a haptic warning system? • What are the remaining functional requirements and specifications of an ICA system? • What DII and DVI will result in optimal driver response? • What technologies show the most promise for the feasible architectures? • What ICA architectures are feasible for meeting the requirements and specifications?

32. What is the optimal timing for collision warning? • Need to continue to systematically determine the ‘too late’ points for various DIIs and DVIs • A goal of zero misses is being used • Driver acceptance is being considered • ‘Too late’ thresholds are being contrasted with known ‘too early’ thresholds, to determine the potential for nuisance alarms

33. What is the optimized braking profile for a haptic warning system? • It is known that severely distracted drivers can have perception reaction times as long as 4 sec, which would make any traditional warning ineffective • A brake assist or full brake system is seen as a possible means of aiding these drivers, since reaction time is eliminated • Three issues are being resolved: • When should the system be activated? • How long should the system remain active? • How much braking authority should be used?

34. Brake Assist Example

35. Savings of 50ft at 35mph

36. What are the remaining functional requirements and specifications of an ICA system? • Need to test between 1 m and 2.5 m on lateral position. • Need to determine effects of various communication system update rates

37. What DII and DVI will result in optimal driver response? • Will continue to conduct evaluation of brake assist and full brake options. • Will continue to test auditory and haptic options.

38. What technologies show the most promise for the feasible architectures? • VTTI will continue to determine technologies that meet the minimal functional requirements • Controller • Positioning • Sensors • Driver Interface • Communications • Computations

39. Controller Technologies • No single interface standard available • Available timing information not accurate enough • Ex.: Eagle controllers report timings to whole seconds • Overhead from 10Hz polling may overload controller • May need to mandate standards for data format/availability • New Advanced Traffic Controllers (ATCs) may address some of these issues

40. Positioning Technologies • Infrastructure based • Radar: costly to cover all lanes • RFID: may require multiple readers per approach • Vehicle based • GPS with INS: high cost to get high accuracy and update rate • RFID in conjunction with odometer: high accuracy at one distance, then decrement remaining distance using odometer

41. Sensor Technologies • Infrastructure • Radar: velocity and deceleration • Vehicle velocity • GPS with INS: velocity • Velocity from vehicle network • Vehicle deceleration • Accelerometer to sense braking • Mechanical sensor on brake pedal

42. Driver Interface Technologies • Infrastructure • Strobes • VMS sign • Intelligent rumble strips • Vehicle • Auditory: tones or voice warning • Haptic: Soft braking (pulses), seat shaker, brake assist, or full braking

43. Communications Technologies • Must be generic to support multiple interfaces • Bi-directional link depending on architecture • DSRC current best choice • Not yet available off-the-shelf • Security issues • Styling issues • Currently simulating with 802.11a hardware and software

44. Computations Technologies • Infrastructure • On board signal controller • Custom DSP or hybrid microcontroller • Must talk to infrastructure components • radar • RFID • DSRC • DII • Easily modified to allow algorithm changes

45. Computations Technologies • Vehicle • Custom DSP or hybrid microcontroller • Must talk to all necessary data sources • vehicle network • DSRC • GPS w/INS • RFID • DVI • Easily modified to allow algorithm changes

46. What ICA architectures are feasible for meeting the requirements and specifications? • VTTI will continue to evaluate the available technologies as suitable to several architectures • Infrastructure only • Mostly infrastructure based with receiver and DVI in vehicle • Mostly vehicle based with transmitter in infrastructure (provides stop bar location and signal phase/timing) • Totally vehicle based with map in vehicle • For stop signed intersections

47. Architecture Example 1. RFID tag reader and in-vehicle warning DSRC transmits signal phase and timing to vehicle DVI in vehicle presents warning and driver stops Odometer updates distance to intersection RFID transmits distance to intersection to vehicle

48. Architecture Example 2. Infrastructure radar and DII DSRC violation warning also transmitted to properly equipped vehicles DII in infrastructure presents warning and driver stops Algorithm calculates violation Radar determines vehicle speed and distance Radar determines vehicle speed and distance

49. Architecture Example 3. In-vehicle positioning system and in-vehicle warning STOP STOP STOP STOP DVI in vehicle presents warning and driver stops Distance and speed calculated to check for violation In-vehicle map query for stop bar location

50. Additional Pre-FOT Testing • Next steps toward an FOT • Passive evaluation of both infrastructure based and cooperative ICA systems • When was the warning issued? • How many false alarms and misses occurred? • What would have been the resulting driver and traffic consequences?