Adaptive Traffic Signal Control Highlights – ITS Michigan and Michigan ITE, February 2012

On February 8, 2012, the ITS Michigan Quarterly meeting was combined with the Michigan Section Institute of Transportation Engineers meeting at the Rackham Conference Center at the University of Michigan campus in Ann Arbor.  Dave Miller, Director of Engineering for Siemens ITS in Austin, Texas, discussed preparation of the infrastructure for connected vehicles.  He noted that there are three classes of on board equipment (OBE) for vehicles.  Class 1 OBE is permanently installed.  Class 2 OBE is carry on after- market equipment,  Class 3 OBE is consumer grade (GPS or Smartphone).  The Connected Vehicle Basic Elements are the vehicle elements (OBE), the stationary infrastructure (roadside equipment), and the “message set” (SAE J2735).  The traffic controller industry estimates that 96,000 of the 307,000 traffic signal controllers currently in place in the United States have the capability to communicate with vehicle on board equipment.  The cost to upgrade the remaining traffic signals is estimated to be $352 million.
Danielle Deneau, Director of Traffic & Safety for the Road Commission for Oakland County, described the SCATS adaptive traffic signal system control used at 675 of the county’s signals.  The Road Commission wanted a system that was cutting edge with real-time traffic signal plan generation.  SCATS uses degree of saturation and volumes to adjust splits, cycle lengths, and offsets.  The system marries signals to the system when required but divorces the signals when traffic patterns change.  SCATS uses a variety of communications systems from 9600 baud phone lines to 900 MHz radios.  The benefits of the systems have been a reduction of traffic crash severity, and reduction in travel times, and a reduction in stopped delay.

Les Sipowski from the City of Ann Arbor described the SCOOT adaptive traffic signal system used in Ann Arbor.  Ann Arbor chose SCOOT because of the city’s familiarity with Eagle/Siemens equipment.  It was an off-the-shelf system used by hundreds of communities and had positive reports from it applications in Toronto, Ontario, Canada.  SCOOT monitors all detectors 4 times per second and reports back to the central computer once per second.  Eight seconds before the new cycle it evaluates conditions and recommends changes based on detector inputs.  The cycle can be adjusted by 4, 8, or 16 seconds, and splits may be extended as needed.  Offsets can be adjusted by 4 seconds.  SCOOT generally keeps saturation levels below 80%.  The Ann Arbor experience is that SCOOT provides fast response with advanced detection.  It has a bus priority package, and it handles pedestrians well.  It responds well to freeway incidents and is very helpful in handling traffic on football Saturdays. For signal phasing, lagging left turns generally work better than leading left turns.

Eric Gannaway, Regional Account Manager for Rhythm Engineering explained the InSynch Adaptive Traffic Signal Technology.  The InSynch adaptive traffic control system uses artificial intelligence to optimize traffic signals at individual intersections and coordinate traffic signals along arterial corridors to reduce traffic congestion.    By reviewing the system’s main hardware and software components, its optimization methodologies and available add-on modules, InSynch overlays existing traffic cabinets and controllers to intelligently and immediately adapt to real-time traffic demand.

InSych’s adaptive technology works in real-world scenarios to intelligently improve traffic flow, thus improving safety and travel time for motorists while also decreasing wasted fuel and harmful emissions.  The essential components of the InSynch system – digital internet protocol (IP) cameras, the InSynch processor and the selected method of connecting to the controller through detector cards or cabling are compatible with existing digital controllers, functioning as an overlay system the simply plugs into the existing traffic cabinet hardware.

Typically installed on the mast arms of traffic signals, the IP digital cameras detect presence like traditional image detection cameras, but also measure occupancy, queue length, and delay every second and communicate that information through an Ethernet connection to the InSynch processor, which resides at the local traffic cabinet.  Based on the information from the cameras showing the real-time traffic demands at the intersection, the InSynch processor determines the priority for service for each approach.  Because the processor is a modern, digital state machine (non-linear and non-sequential) the system is able to serve traffic demand without being inhibited by pre-determined cycles or splits.

The InSynch processor requests a green signal for the state that is most appropriate to serve by inputting the appropriate calls into the signal controller, which runs in free mode to allow for the acceptance of the detector calls.  InSych’s artificial intelligence in comprised of a local optimization algorithm for each intersection and global coordination between all the intersections on a corridor.  The intelligent actuation and global coordination work in tandem to reduce stops and delay along the corridor.

Mark Hudgins, Adaptive Systems Project Manager for Siemens, described the use of ACS Lite for adaptive traffic signal control.  ACS Lite uses time of day patterns and optimizes splits and offsets.  It does not change signal cycle length.  ACS-Lite employs the concept that a TOD schedule is an appropriate way to manage traffic demand over the day and by days of the week. Within the context of a TOD schedule, ACS-Lite will adapt the particular plans that are implemented at each time of day based on the overall performance of that plan for the similar previous day.

This approach to adaptive behavior uses the traditional traffic engineering assumption that average behavior of traffic on, for example, Tuesday at 3 P.M., is roughly the same on every Tuesday at 3 P.M., but drifts slowly with long-term changes in population, construction, new routes, etc.

If the performance of the baseline plan is determined to be improvable by changing cycle, splits, or offsets, then those changes will be made to the “optimized” plan stored in ACS-Lite and downloaded to the local controllers for use on the next day. The goal of being appropriately adaptive at this level is the maintenance of the timing plan over long periods of time to address the typical degradation of plan effectiveness (e.g., 4% worse per year) and replace the very expensive task of re-timing signals on a periodic basis.

The next level of adaptivity used by ACS-Lite is on-line modification of the TOD plan parameters as the plan is running. With the assumption that the baseline optimized TOD plan is a good starting point, ACS-Lite will adapt the cycle, split, and offsets of the plan within some neighborhood of the baseline settings over the plan’s intended implementation duration. ACS-Lite may also adapt the start and end time of the plan from the baseline TOD schedule according to the current conditions, considering the effectiveness of the new plan versus the one that is currently running. ACS-Lite also identifies and selects the best strategy to transition between timing plans.