1. HIOCC Developments SEA HA

2. overview problems with HIOCC developments to address problems simulation using Individual Vehicle data demonstration of simulator tool

3. HIOCC problems Key: false alerts due to HGVs long clearance time under low flow poor detection during night Additional: choosing appropriate parameter settings control of triggering speed

4. understanding false alarms / detection speed-length ambiguity since using only occupancy data  2.0s occupancy = 4m car at 6.7mph  2.0s occupancy = 16m HGV at 20.1mph no synchronisation between arrival of vehicle at detector and start of instantaneous occupancy measuring interval  for occupancy times 2.0s-2.8s, only some vehicles will trigger  2.9s needed to guarantee alert  slower detection speed

5. occupancy time vs. speed

6. detection uncertainty

7. detection regions red = alert possible alert green = no alert

8. detection regions red = alert possible alert green = no alert length distribution (M25 site 4762, 4 May 2002 11:19-18:28)

9. clearance under low flow caused by suspension of smoothed occupancy calculation feature intended to prevent alert fluctuation in stop-go traffic, but cannot distinguish from case of no vehicles after (false) alert in very low flow  alert does not clear on otherwise empty lane

10. development of HIOCC2 make use of individual vehicle speed data  trap and suppress false alerts raised by HIOCC doesn’t modify basic HIOCC algorithm  maintain detection performance  tried and tested HIOCC2 easily switched out to revert to standard HIOCC  robustness to missing/outlier speed data  robustness to single loop failure

11. HIOCC2 detection regions

12. HIOCC2 architecture

13. HIOCC2 key points Watchdog suppresses HIOCC alert if speed > threshold  removes false alerts smoothed speed for clearance  clear only after n vehicles passed  tolerance to outlier speed  force HIOCC to clear if smoothed occupancy calculation suspended Watchdog output forced high if loop fails  revert to HIOCC data synchronisation to align occupancy and speed data

14. HIOCC2 issues algorithm introduces 3 second delay  robust detection in case of vehicle stopping on first loop but not reaching second  to ensure suppression of long vehicle false alarms  potential for using the time for pre-processing to improve HIOCC detection significant reduction in false alarm rate detection performance of HIOCC not changed threshold speed controllable

15. improving HIOCC detection aim to remove the uncertain detection region through pre-processing occupancy overcome the problem of synchronisation features :  raise detection speed  self-contained, no modification to HIOCC  imposes 1 second delay 6kmh (1.68m)

16. pre-processing detection regions

17. simulation simulation of HIOCC based on logged IVD data software tool developed to simulate HIOCC/HIOCC2  step through second-by-second and view/plot internal variables, inputs, outputs  measure overall performance  automatically test different parameter combinations  read IVD logged from different Outstations: [Peek, Golden River, Siemens] length/speed headway length/speed

19. benefits open up ‘black-box’ to understand / visualise operation in different traffic conditions test algorithm on specifically generated test examples and situations  ‘what if’ testing rapidly and automatically try out different parameters

20. dataset from M25 dataset collated from M25 Individual Vehicle Data (May 2002) examples of incident / non-incident / HIOCC false alerts for use in the simulator  different location types: hill, junction, normal  different traffic flows

21. potential development Get as much as possible from existing equipment. aim to improve control and allow higher triggering speed  iterate HIOCC at faster rate (e.g. every 0.5s)

22. summary developments have been carried out that:  address problems of HIOCC  keep the underlying HIOCC algorithm, which is always used as a failsafe fallback simulation tool now available

23. improved

24. current