This paper describes how Atkins’ calibration engineers have identified improvements to the calibration process for smart motorway (SM) schemes through their experience on many schemes in the last five years. The M62 J25-30 calibration and M25 SM optimisation schemes are used to provide examples of lessons learned.
How smart motorways work
The photo in Figure 1 shows the key items of physical infrastructure on a SM all lane running (ALR) scheme. The most obvious things to the user are the signals either road side or located on gantries showing variable mandatory speed limits (VMSLs) and an additional permanent running lane which previously was the hard shoulder. Other notable features are CCTV, message signs to inform drivers, emergency areas for broken down vehicles and importantly, detectors to monitor the traffic variables: speed, flow and occupancy.
Dynamic Hard Shoulder (DHS) was used prior to the introduction of ALR; this provides an additional running lane during peak periods only but requires significant additional infrastructure for monitoring the hard shoulder; the monitoring and opening process is also inefficient. DHS schemes are still present on the strategic road network however all new SM schemes are designed and built either as ALR or Controlled Motorway with a hard shoulder.
SM has three modes of operation; these are actually algorithms which are applied separately and dynamically, in so much that each mode of operation can be applied in different locations along the scheme at the same time. The mode of operation is selected using data from traffic detectors, either mounted above the near side of the carriageway or buried in the road surface which measure speed, flow and occupancy on in each lane at approximately 500 metre intervals.
The modes of operation are:
Congestion Management (CM): CM uses 50mph and 60mph VMSLs for managing increases in flow that can lead to flow breakdown (congestion). In so doing it delays the onset of flow breakdown and improves the recovery. These speed limits reduce the speed differential between lanes, giving a more laminar flow and making dangerous manoeuvres less likely. They also have the potential to improve vehicle emissions because they encourage drivers to maintain a constant steady speed, rather than accelerating and braking. If stop-start congestion can be avoided then the benefits could be even greater.
Dynamic Hard Shoulder (DHS): Operators in the control centre are prompted by increasing flows to perform an opening sequence, using continuous CCTV coverage dedicated to checking the hard shoulder on each link is clear before opening it. A speed limit is always displayed when the hard shoulder is open, 60mph or lower.
Queue Protection (QP): 40mph mandatory speed limits are set in response to high occupancy on any given traffic detector site in the scheme, typically 90% occupancy over a period of 2 seconds. 40mph signals are set for safety management, not congestion management, to protect queuing traffic.
The key SM benefits are:
Capacity: The additional lane either provided permanently by ALR or dynamically through DHS quite simply increases the capacity of the road, meaning higher flows can be accommodated preventing congestion during peak periods. (This is at least the case for as long as induced traffic or general traffic growth doesn’t exceed the new capacity available.)
Journey Times: The CM VMSLs tend to make journey times more reliable, and reduce them in peak periods.
Safety: The current safety objective for SM schemes is to maintain the high safety standards from pre-scheme but many schemes have shown improvements in collision rates. For example, a safety performance review of the first two SMALR schemes (M25 J23-27 and J5-6) showed a combined reduction in the overall collision rate of 17% after two years. This equates to a combined reduction of 7%, after background trends of reducing collision rates have been taken into account.
What is calibration?
Calibration or optimisation is the process of formulating, monitoring and fine-tuning the correct thresholds for entering into the system for operating each algorithm to ensure the optimum benefits are realised.
This is necessary because the thresholds for setting VMSLs must be tailored to the particular location. The traffic characteristics are influenced by topology, road layouts, flows, number of HGVs and origin-destination of vehicles.
The CM algorithm uses speed and flow thresholds which trigger the 60mph and then 50mph signals to be displayed according to pointers from each detector location to signals on gantries upstream in the traffic flow. These thresholds need to be calculated, tested and entered into the site data for each individual detector location. The aim is to identify locations which regularly cause congestion or flow breakdown (seedpoint locations). A 50mph VMSL should be displayed upstream of these locations a few minutes before the congestion is likely to occur. 50mph is the optimum speed for improving traffic flows; the flow becomes more laminar which can delay or prevent the onset of congestion. The algorithm also sets 60mph VMSLs as a ‘lead-in’ to the 50mph signals.
The most significant part of the calibration exercise comes in identifying the flows to activate and deactivate the 60mph and 50mph VMSLs.
The CM algorithm does not have the ability to self-tune or the ability to use multiple CM thresholds per site. Therefore, the thresholds must necessarily be a compromise, for example if the system is tuned for the AM peak it may not work optimally in the PM peak. The only way to achieve this compromise is by analysing the traffic data to ensure the system will give the best performance overall.
The DHS threshold is known as the ‘link flow state threshold’ (LFS); there is a standard to set this in accordance with the thresholds set for the CM algorithm. However, in reality fine-tuning the LFS for opening and closing the hard shoulder is done by liaising closely with the operators. It is more important to choose values that work for them and their operational judgment. There is a risk that if operators are not consulted they override the LFS without a full appreciation of what affect it might have on the traffic. This is quite important when opening the hard shoulder, to ensure that enough time is allowed for the opening process before congestion occurs. It is vital when closing the hard shoulder because the effect could easily be moving four free flowing lanes of traffic into three lanes of instant congestion!
Queue protection thresholds are relatively standard across all motorways in the Highways England strategic road network and do not require optimisation.
Standard calibration and optimisation process
There are several steps to the standard calibration process as specified in MCH2584 ‘Guidance for the calibration and optimisation of smart motorway systems’, as shown in Figure 2.
In order to calculate the various thresholds a representative sample of high quality data from a period not during works, construction or school holidays, etc. is required.
The CM VMSLs are triggered using flow and speed thresholds, which, in the standard approach, are configured off-line using pre-scheme data before the scheme opens and then calibrated and optimised in the post-opening period.
When a 3-lane motorway is converted to 4-lanes (either for ALR or DHS), there is no existing data for the 4-lane scenario which can be used for the off-line configuration of thresholds. The only information available relates to the 3-lane situation in the pre-works period. Current practice, as implemented in the Highways England’s SM CALO tool, is to calculate the 4-lane thresholds by ‘scaling up’ from the 3-lane pre-works flows, i.e.:
- Identify the flow at which congestion generally occurs with 3-lane
- With 3-lanes, the ‘Rising 50’ threshold would be set to slightly less than this flo
- Multiply by 1.3 to obtain the ‘Rising 50’ threshold for 4-lanes
The pre-construction data is used to calculate and test the thresholds offline. This happens at least six months before a scheme is ready to go live, which gives time to enter the values into the site data in the control centre.
For added confidence, it is usual to ‘preview’ the system performance the week before switch-on. This means that new data from the new detection sites which will operate the system is used to check how the system will behave. It is particularly important to do this as road layouts and traffic flows could have changed since before construction.
Once the scheme is complete and the technology has been commissioned, the algorithms are enabled.
Following this process monitoring and calibration occurs for a period of up to two months. Any observations from this process can be addressed by small ‘tweaks’ to calibrate the thresholds on individual detection locations.
Monitoring and optimisation:
The calibration engineers are then responsible for further monitoring and optimisation, for a period of 12 months after switch on. This is very useful as often the additional capacity gained by use of the hard shoulder means it takes quite a while for congestion to start to occur.
Process improvements identified through experience
Threshold configuration (Fig 3)
Atkins used the above method (via SM CALO) to create the 4-lane thresholds on M62 J25-30 SM. Figure 4 and Figure 5 are before and after space-time plots of the M62 scheme.
The plots show time (horizontal axis) and distance (vertical) with junction numbers on the right and detection / signals on the left. Pale grey areas show low speeds; coloured horizontal bars show VMSLs displayed on gantries after the scheme has been implemented (green for 60mph, orange for 50mph and yellow for 40mph).
On this scheme, there was a key bottleneck in the pre-works 3-lane scenario at Junction 27 eastbound. During the first month after opening, the extra capacity provided by DHS had removed this congestion. However, the CM algorithm was setting 50mph VMSLs on a regular basis, despite the lack of congestion. This means drivers were being slowed to 50mph for up to 50 minutes for no reason; this reduces the early realisation of benefits and increases the risk of poor driver perception.
The same thing occurred on more than 60% of days in the System Monitoring period. Atkins increased the thresholds in the next site data load to prevent this inappropriate setting of signals, but there is a 6-week delay before this new site data is installed. When traffic flows began to increase, congestion did start to occur in this location, a System Monitoring exercise was performed to identify the most appropriate flow to trigger 50mph and 60mph CM VMSLs, based on the actual traffic conditions at the time.
The lesson learned from this is that there was no benefit (and actually a detriment) to calculating 4-lane thresholds using the ‘Multiply by 1.3’ rule. It would have been better to configure the thresholds to high initial values, removing the risk of setting inappropriate 50mph VMSLs. 40mph VMSLs would still provide queue protection meaning safety would not be compromised. The CM algorithm would still set 60mph and 50mph VMSLs when speeds on the motorway drop to 50mph and 30mph, respectively. So drivers would still perceive the system to be in operation.
Having learned this lesson on one of our first calibration exercises, we introduced an improvement on all subsequent calibrations with the agreement of all stakeholders; we now configure the thresholds only once the scheme has gone live. This significantly reduces costs as well as allowing us to get thresholds ‘right first time’, improving road user experience.
Monitoring and optimisation (Fig 6)
Atkins have optimised a large number of schemes including all sections of the M25. As part of this process, we again identified many locations where 50mph VMSLs were being displayed unnecessarily because the thresholds were too low. By increasing these thresholds and ‘future-proofing’ to prevent the same issue recurring, we have improved journey times and road user experience; in several places the number of complaints to Highways England’s Information Line was found to decrease after optimisation.
During the optimisation and monitoring the calibration engineers were frequently present within the RCC and routinely engaged with the Highways England Regional Control Centre (RCC) Operational Teams and Calibration Managers.
The key stakeholder engagement created an environment for better communications and knowledge sharing. Importantly a presence and engagement also allowed for CM threshold changes to be discussed and agreed prior to site data loads. This has recently become more important as Highways England set up a team of their own Calibration Managers to take responsibility for optimisation of smart motorways once the scheme’s calibration is complete. We have worked closely with the Calibration Managers to ensure they can benefit from our lessons learned and experience.
This had a number of advantages, for example, on the M62 J25-30 SM scheme fine tuning for best overall performance was discussed and agreed prior to site data loads. The Operations teams were also consulted on the LFS values (DHS opening and closing prompts) to minimise the risk of leaving the hard shoulder open when not required, capture required time allowances for DHS opening and closing processes and accommodate RCC operations including shift changes.
Currently there is no specific requirement for calibration engineers to be present at the RCCs; however we have identified real benefits on our schemes of being able to share knowledge and improve understanding in a very efficient way.
Calibration of SM schemes is crucial to achieve the optimum benefits in terms of capacity, journey times and safety. This paper has provided an overview of the SM optimisation process and identified some positive benefits following changes to the standard processes.
Setting CM flow thresholds high prior to switch on has been adopted on other schemes and has been a benefit by preventing inappropriate CM VMSLs and allowing for thresholds to be calculated based on actual 4-lane data. Also allowing a short period after opening for the traffic to settle has helped identify congestion seedpoints following changes to the topography and lane geometry.
Actively engaging RCC teams has allowed for a more holistic approach in calibrating and optimising SM schemes. Highways England have also recently restructured by creating a calibration and optimisation team and we have been working closely with them in all our SM calibrations.
The revised approaches ultimately benefit road users by minimising unnecessary VMSLs and have improved their perception of SM. ◆