The Hajj is an annual pilgrimage that attracts millions of pilgrims to the city of Makkah, Saudi Arabia, each year.  The Hajj occurs from the 7th to the 13th day of Dhu al-Hijjah, and involves a series of rituals conducted by the pilgrims in Makkah and several surrounding sites.
The pilgrimage route takes the pilgrim to the Masji al-Haram in Makkah, and then on to the town of Mina, the Plains of Arafat, and Muzdalifah, before heading back to Mina and then back to Makkah.  To transport pilgrims during the Hajj, approximately 30,000 motor coaches are used.  Due to the large numbers of pilgrims and motor coaches involved, the safe and efficient transportation of these pilgrims through the various locations of the Hajj is an enormously challenging transportation problem.

Evidence from previous Hajj’s suggests that one segment of the Hajj, the journey from Arafat to Muzdalifah (a distance of approximately 5 km), can take as long as 5 to 6 hours as a result of extremely congested roadways.  Pilgrims do not leave Arafat until after sun down, and are supposed to arrive at Muzdalifah before midnight.  Given the necessity for pilgrims to complete this journey within a roughly six hour period, this level of congestion represents a significant transportation bottleneck within the Hajj.
For the 2009 Hajj, a portion of the buses used to transport pilgrims from Arafat to Muzdalifah were outfitted with Radio Frequency Identification (RFID) readers and Global Positioning System (GPS) data loggers in order to gain a better understanding of bus operations.  In tandem, a new operational plan was tried in which centrally coordinated buses made repeat trips between the two locations, rather than having pilgrims take their own tour bus as had been done in the past.

The data obtained from the RFID readers and GPS data loggers provided an opportunity to evaluate the new operational plan, so that lessons learned can be extended to future operations of the Hajj Pilgrimage.

The data obtained indicated that the coordinated operational plan was extremely successful.  Significant congestion along the roadway between Arafat and Muzdalifah was essentially eliminated, and the average travel speed was 64 km/h.  Over half the trips were completed by 8:30 p.m. and approximately 95 percent of trips were completed by midnight.  It was also found that travel time  remained relatively consistent throughout the day.

Improving and optimizing transit service requires data, and traditionally, collecting data on transit system performance has been done manually. This is problematic because the significant cost and complexity of manual transit data collection has forced many agencies to make do with limited datasets for planning, operating and evaluating their networks.  Recently, many transit agencies have begun to implement automatic transit data collection via Automatic Vehicle Location (AVL) and Automatic Passenger Counting (APC) systems.  These systems allow the collection of large, detailed datasets of transit operations, and provide an opportunity to evaluate and optimize transit operations using methods that were infeasible in the previous data-poor environment based on manual data collection.

This research developed a methodology to utilize data collected by typical AVL/APC system installations in order to (a) develop advanced performance measures to quantify schedule adherence and (b) automatically determine the causes of poor schedule adherence.  The methodology addresses the difficulty that many small to medium sized transit agencies have in utilizing the data being collected by proposing a methodology that can be automated, thereby reducing resource and expertise requirements and allowing the data to be more effectively utilized.

Traditional schedule adherence performance measures identify how frequently a bus is “on-time” along a given route or at a given stop.  Because AVL/APC systems collect data on passenger activity as well as arrival and departure times, it is possible to improve upon traditional performance measures by looking at schedule adherence from a passenger perspective and seeing how frequently a passenger experiences a bus that is “not on-time” along a route or at a stop.  This frequently results in a vastly different list of routes and stops that need improvement, because stops where buses are most frequently “not on-time” do not always coincide with the stops where passenger activity is the highest. 

The causes of poor schedule adherence are identified based on patterns in the data and categorized as travel time causes (not enough/too much travel time in the preceding segment), dwell time causes (the bus is dwelling longer/shorter than scheduled at the previous/current timepoint), and upstream causes (the bus was already late/early at the upstream timepoint, indicating the cause is occurring upstream).  These causes, while general in nature, relate back to “real-world” reasons that a bus is late or early.  Correcting these “real-world” reasons provides a way to address and correct areas of poor schedule adherence.

The causes are presented as cause statistics, which represent the percentage of time that each possible cause was found to be a factor at a given stop when a bus is late or early.

In addition, a visualization aid is developed to show the scheduled trajectory between a given pair of timepoints and the true trajectories of each observed trip between the same pair of timepoints.  The trajectories are plotted using the most detailed resolution of the AVL/APC system data possible, which means that information may be provided even between timepoints.  Combining all the true trajectories into an average true trajectory permits identification of where delays are occurring (such as at intermediate stops or intersections); where the bus is going faster or slower than scheduled (by comparing the slope of the average and scheduled trajectories); variation in trip travel times (by looking at the spread of the trips); etc.  This form of diagram provides additional data that can be used to confirm the cause statistics and help develop an appropriate strategy to address any issues that had been previously identified.

The ultimate output of the proposed methodology includes the following:

1. A ranked list of routes by direction (for a given time period) that identifies routes with the poorest schedule adherence performance.
2. Performance measures within any given route, direction, and time period that identify which timepoints are contributing most to poor schedule adherence.
3. Statistics indicating identified causes of poor schedule adherence at individual timepoints.
4. A visualization aid to be used in conjunction with the cause statistics generated in Step 3 in order to develop an effective strategy for improving schedule adherence issues.

With this information, transit agencies will be able to act proactively to improve their transit system, rather than wait until they discover problems on their own or hear complaints from passengers and drivers.

The purpose of this study was to quantify the safety and traffic flow impacts of a candidate Variable Speed Limit Sign (VSLS) control strategy under different traffic scenarios for an urban freeway section. Four traffic scenarios were modelled, each under a different condition of recurrent or non-recurrent congestion. The effects of the VSLS control strategy on safety and system delay were determined using a microscopic simulation model (PARAMICS) combined with a categorical (log-linear) crash potential model. Impacts to safety were measured by changes in crash potential, calculated through the crash potential model from simulated loop detector data. Impacts to traffic flow were measured by travel time data output by the simulation.

The VSLS impact analyses were performed on three traffic scenarios of varying levels of congestion – peak, near-peak, and off-peak, and on one scenario of non-recurrent congestion involving an incident of 15 minute duration. VSLS impact was quantified in terms of the relative changes in safety (crash potential) and vehicle travel times before and after the implementation of the VSLS control strategy. A positive change in relative safety benefit represents a decrease in crash potential while a positive change in travel time represents an increase in average travel time per vehicle. The most desirable outcomes for the VSLS impact would be a large positive relative safety benefit associated with a decrease in travel time. The following table summarizes the overall average safety and travel time impacts for each scenario, including the average percent time each speed limit was displayed during the simulation period.

Conclusions
The most desirable outcomes for a VSLS impact would be a large decrease in crash potential associated with a decrease in travel time. Overall the results provide no clear indication that the implementation of a VSLS system under the current control algorithm would positively impact safety and travel efficiency measures for all traffic scenarios. However, the analyses of the VSLS impacts under this particular control algorithm do provide insight and evidence that suggest the following:

• Traffic scenarios experiencing higher congestion are more likely to benefit from a VSLS system in terms of higher positive relative safety benefits and less negative travel time impact than traffic scenarios with less congestion. These benefits appear to occur, at least in part, as a result of the reduction in the frequency and severity of shockwaves in the congested traffic (i.e. damping of the stop and go oscillations).
• The most congested locations or locations which trigger speed limit decrements are more likely to experience positive relative safety benefits with less impact to travel time.
• For less congested conditions, stations upstream of VSLS response zones are more likely to experience negative relative safety benefits.
• Vehicles making longer trips are more likely to experience negative travel time impacts under the current VSLS control algorithm than vehicles making shorter trips.

The most desirable results (both positive safety and positive travel time impacts) were usually observed for moderately congested scenarios during which the VSLS response exhibited frequent speed limit decrements and frequency recoveries. The least desirable results were usually observed under conditions that caused prolonged speed limit reductions and thus lower freeway speeds than would have been observed without VSLS. This suggests that the tested VSLS control algorithm was able to provide large safety benefits with no significant travel time penalty, but only for a limited range of traffic conditions. It is suspected that by modifying the parameters within the current algorithm (e.g. occupancy threshold, volume threshold, response zone requirements, etc.), the VSLS may be able to operate effectively over a wider range of traffic conditions and provide more consistent safety and travel time benefits. It is recommended that alternative VSLS control algorithms be explored and it is suggested that the evaluation framework used in this study is an effective tool for optimizing the algorithm structure and parameter values.

The previous VSLS evaluation study used a micro-simulation (PARAMICS) model to assess the impacts of variable speed limits. This study assumed that driver reaction to VSLS posted speed limits (particularly those lower than the existing static speed limit of 100 km/h) would be the same as driver reaction to the existing static posted speed limit. This means that individual driver compliance with the posted speed limit would be the same for VSLS posted speed limits as for fixed statics speed limits. Thus, a driver that elects to exceed the posted static speed limit by 10% (i.e. drive at 110 km/h for a posted speed limit of 100 km/h) would drive at 88 km/h for a posted speed limit of 80 km/h and 66 km/h for a posted speed limit of 60 km/h.

However, other driver reactions to VSLS posted speed limits may be possible. For example, fewer drivers may comply with the speed limit because they know the road can be safely travelled at higher speeds during other time periods. Alternatively, it is possible that drivers compliance with the speed limit will increase because they may feel that a speed limit tailored to actual conditions is more relevant than a static speed limit that is applied “by default” across the province under all weather and traffic conditions.

Since variable speed limits are usually only lowered during periods of congestion, diver speed choices will often be governed not by the posted speed limit, but by the traffic conditions themselves. However, during some periods, such as when congestion is building or dissipating, traffic conditions may not be the governing factor and driver reaction to speed limits will play an important factor in the operation of the VSLS system. Therefore, the output of any micro-simulation analysis of VSLS control strategies is expected to be at least partially dependant on the assumptions that are made with respect to driver reactions and compliance to the variable speed limits.

The purpose of this study was to investigate the sensitivity of VSLS safety and traffic operations (measured in terms of travel time) performance to speed limit compliance.

Figure 1: VSLS performance sensitivity to speed compliance
(Average of 10 simulation runs for Peak traffic scenario)

A number of observations can be made on the basis of the result in Figure 1:

1. Results differ from those obtained in the previous VSLS study likely because of the use of different versions of PARAMICS. The previous study used model version 5.1. The current study used model version 6.4. These two model versions provide different safety and travel time measures of performance for the Non-VSLS scenario suggesting that changes made to the simulation software have had some impact on driver behaviour. This makes direct comparisons between the previous and current VSLS simulation studies difficult.
2. Compliance level has a very significant impact on VSLS safety performance. As expected, benefits increase (though non-linearly) as compliance level increases. The largest increase in safety occurs for the change from Low to Moderate compliance. Increases in safety become smaller as compliance increases. Nevertheless, safety continues to improve for all increases in speed limit compliance.
3. Compliance level also has an influence on VSLS travel time impacts. As expected, travel times increase as compliance level increases. The impact on travel time is relatively modest for the Low, Moderate and High compliance scenarios as the change in travel time with changes in level of compliance is small. However, for the Very High compliance scenario, the increase in travel time is very large. This result might suggest that there is little incentive to implement VSLS in a way that attempts to achieve very high speed limit compliance, as the additional safety benefit is small and the cost in terms of increased travel time is very large. However, this result is unexpected and somewhat counter-intuitive. It was speculated that these results reflected characteristics of the VSLS strategy evaluated (and the VSLS parameter values) rather than VSLS per se. More specifically, it was hypothesized that under the Very High compliance level, the existing VSLS strategy would result in a lowering of the speed limit to 60 km/h, but would not subsequently increase the speed limit when expected (i.e. when measured loop detector occupancy was less than 15% for three consecutive 20-second intervals). This hypothesis was confirmed through a detailed examination of the simulation results.

As a result of these observations, additional simulation runs were conducted to determine the performance of VSLS under the Very High compliance scenario when the occupancy threshold for incrementing speed limits was changed from 15% to 20% and the speed limit threshold for decreasing speeds was reduced by 10 km/h (i.e. 80 km/h was changed to 70 km/h; and 60 km/h was changed to 50 km/h).

Figure 2 provides the safety and travel time impacts for these three Very High compliance scenarios averaged over 10 simulation runs. These results indicate that safety and travel time impacts are dependent on the parameter values chosen for the VSLS strategy. This finding is consistent with the results of the previous VSLS study. However, this finding also suggests that the most appropriate (optimal) set of parameter values is a function of the level of speed limit compliance exhibited by drivers. Furthermore, these results suggest that the safety benefits associated with the Very High compliance scenario as indicated in Figure 1 are modestly under-estimated and the travel time increases are significantly over-estimated.

Figure 2: Impact of parameters on VSLS performance for the Very High compliance scenario

Conclusions:

1. VSLS impacts, in terms of safety and travel times, are quite sensitive to the level of speed compliance. The safety benefits of VSLS under the Very High compliance scenario are more than 4 times the benefits obtained under the Low compliance scenario.
2. Safety benefits of VSLS increase with increasing speed limit compliance.
3. Travel time penalties (i.e. increases) that result from VSLS also increase with increasing speed limit compliance.

The provision of transit vehicle priority is often motivated by opportunities to reduce person-delay within the transportation network, increase transit reliability and speed, reduce transit operating costs, and/or encourage transit use due to the environmental and social benefits often associated with transit. Within the freeway environment, one method of providing transit priority is via transit “pass-through” lanes at interchanges. “Pass-through” lanes allow a transit vehicle to exit the freeway at an interchange, cross the intersecting arterial road, and re-enter the freeway.

Since limited resources preclude the Ministry of Transportation of Ontario from implementing transit “pass-through” lanes at all interchanges, it is valuable to have a method to evaluate and rank potential locations. This research develops a warrant methodology that can be used by practitioners to aid in deciding whether construction of a transit “pass-through” lane at a given interchange is justified, and to provide a method for prioritizing candidate locations. The warrant methodology provides an objective and consistent decision making method, reduces the effort required for practitioners to assess the effectiveness of a “pass-through” treatment at a given interchange, and helps ensure that limited resources are directed towards interchanges which are expected to experience the greatest benefit per dollar spent.

The warrant procedure is based on an analytical methodology that estimates the benefits and costs of a proposed transit “pass-through” lane. The benefits are compared to the costs in the form of a benefit/cost ratio. A proposed transit “pass-through” lane meets the minimum requirements of the warrant when the benefits/cost ration exceeds an agency defined threshold. The benefit/cost ratio is also used to prioritize interchanges which are expected to receive the greatest benefit per dollar spent.
The benefits themselves are calculated by finding the total time savings for passengers and the total time savings for buses throughout the day. These time savings are multiplied by conversion factors to account for the inherent value of passenger’s time that is saved due to the transit “pass-through” lane, the savings to transit agencies due to reduced bus operating times, and the benefit due to induced transit demand created by making transit more competitive relative to personal vehicles. The general benefit calculation procedure is indicated in Figure 1.

The costs are calculated based on the annualized construction cost over the service life of the transit “pass-through” lane, and the annual maintenance costs. The general cost calculation procedure is indicated in Figure 2.

In addition to a theoretical framework, the warrant methodology was implemented as a Microsoft Excel spreadsheet tool in order to minimize the effort and time required by practitioners to use the warrant.

The warrant analysis was demonstrated through application to two interchanges at which transit “pass-through” lanes have already been constructed. The standard warrant methodology (which ignores changes in future travel time and transit profiles) was demonstrated for both the Highway 401 EB/Avenue Road and the Highway 403 EB/Erin Mills Parkway interchanges. An advanced version of the warrant methodology (which allows the user to specify changes in future travel time and transit profiles) was also demonstrated at the Highway 403 EB/Erin Mills Parkway interchange.

It was found that freeway speeds have a significant influence on the results of the warrant analysis. If freeway speeds are generally high throughout the day, a transit “pass-through” lane is unlikely to meet the warrant requirements, as was the case at Highway 403 EB/Erin Mills. Lane group volumes at the signalized intersection of the off-ramp have a smaller effect on the outcome of the warrant, unless volumes approach or exceed capacity. The transit schedule is also important, as travel time benefits are only accrued during periods in which transit vehicles pass through the interchange. Therefore, the key periods for the warrant to analyse should include times when: (a) there is significant freeway congestion; (b) there are high-volumes on the transit “pass-through” lane group; or (c) there are notable transit volumes.

In 2005 Grand River Transit (GRT) initiated the iXpress express bus service within the Region of Waterloo. Over the following 4 years, GRT initiated a number of supporting technologies and programs including transit signal priority, onboard automatic vehicle location system (AVLS) and automatic passenger counting system (APCS), a web-based trip planner, pedestrian and cycling accessibility improvements, and an individualized marketing program. Dr. Hellinga led a team responsible for developing and implementing appropriate evaluation plans to measure the separate and combined impacts of these project elements, particularly in terms of green house gas emissions. These evaluation activities have made use of a wide range of methods including micro-simulation, statistical modeling, deterministic and stochastic queuing, and statistical inferences from survey data. Data sources included revealed preference and stated preference surveys; on board passenger surveys; non-rider telephone surveys; household surveys; automatically collected vehicle location and passenger load data; fuel consumption data; web usage data; etc.

Final Report

In 2006, the City of Mississauga conducted a pilot deployment of Smart Transit Vehicle applications in order to validate potential benefits. The pilot deployment consisted of (1) on board smart transit vehicle technologies, and (2) transit signal priority measures.

A total of 30 buses were equipped with Automatic Vehicle Location System (AVLS) and Automatic Passenger Counting System (APCS); and (2) . These buses were assigned primarily to Routes 19 and 19A that service Hurontario Street between the Shoppers World terminal in Brampton and the Port Credit GO passenger rail station in Mississauga.

Transit Signal Priority (TSP) was deployed at 14 intersections along the Hurontario (Hwy 10) corridor. The deployment was arranged in two groups – one group to the north and one group to the south of Hwy 401. The TSP locations are denoted by the solid red circles on the map.

The purpose of this study was to evaluate the impact that the AVLS and TSP had on schedule adherence of the route and the variability of bus travel times. The study made use of data collected via the on-board AVLS and APCS to determine the impacts.

The study found:

• Schedule deviation was reduced by 4.6% through the use of AVLS alone, and by 5.2% through the use of AVLS and TSP combined.
• TSP appears to have a greater impact on reducing the variability of travel times than in reducing the average travel times of transit vehicles. The impacts of TSP on reducing variability of transit vehicle travel times seem to vary significantly, in some cases increasing the variability and in other reducing the variability by as much as 50%.