Learning Outcomes

Emerging Trends
CIVE70015 Traffic Engineering By the end of this session you will be able to:

▶ Develop a comprehension of signal priority for buses

Dr Fangce Guo ▶ Review the current car following models and challenges for AVs/CAVs
[email protected]
▶ Compare different capacity management strategies for AVs/CAVs
▶ Discuss the machine learning based applications in traffic engineering
12 December 2023

1 / 33 2 / 33

Lecture outline Section 1

▶ Signal priority for buses

∗ Introduction
∗ Coordination of traffic signals
▶ Car following model for CAVs
∗ IDM Signal priority for buses
∗ ACC/CACC
▶ Capacity management (RoW) in the mixed flow of CAVs and HDVs
∗ Road capacity
∗ Passenger capacity
▶ Machine learning applications in traffic engineering

3 / 33 4 / 33
Introduction Introduction (2)
Bus regularity and punctuality
▶ Punctuality - measure showing the percentage of buses arriving on time, taking account
Objectives for Providing Bus Priority of acceptable tolerance
▶ Economic benefit
▶ Regularity - the measure showing derivation in the actual bus headways from the
∗ More efficient use of space than private cars
▶ More passengers transported in one bus scheduled headway
▶ Average delay reduced ▶ Both are most important factors determining passengers’ perception of service
∗ Total value of time gained by bus/taxi ▶ Irregularity affects passengers waiting at bus stops The mean of the waiting time w spent
passengers and crews outweighs that lost by by passengers at a bus stop is given by
users of private vehicles
▶ Reduction of bus journey time h
E (w ) = (1 + Ch2 ) (1)
∗ Reduction of bus journey time: Increase 2
ridership
∗ Saving running time of buses: More where h is the mean bus interarrival time (headway) and Ch is the coefficient of variation
cost-effective and increase capacity of bus headways. This shows the importance of maintaining a regular bus service in
▶ Bus regularity and punctuality order to manage passenger waiting times.
▶ Irregularity affects both passengers on board the bus and the amount of layover time for
Distributions of bus running times: with vs without priority
the bus and crew
5 / 33 6 / 33

Bus priority measures Signal timings and control

▶ Methods for Fixed-Time Control: the TRRL TRANSYT method

Passive approaches
▶ Methods for Traffic-Responsive Control: the TRRL SCOOT method
▶ Bus lanes
These methods take into account the average behaviour of all vehicles to determine
▶ Calculation of off-line timing plans with consideration of buses suitable offsets for each pair of adjacent junctions, making compromises as appropriate
Active approaches on the basis of the relative benefits of alternative adjustments.
▶ Headway management ▶ Bus Priority
▶ Adaptive signal control with bus priority features ∗ Consideration of buses in signal timing design
∗ Approaches:
▶ Require real time bus detection, by detectors, GPS, etc 1) Consider different sizes of vehicles when calculating performance index (e.g. delay,
▶ More effective than ‘passive’ appraoches passenger-car-unit)
2) Model buses explicitly

7 / 33 8 / 33
Signal timings and control (2) Signal timings and control (3)
Signal timing measure of performance:
Fact: The mean passenger occupancy of buses > private vehicles using the same road space
X
D= qd (2)
Actions:
→ Introduce effects of ‘buses’ into formulation where, q is the mean arrival rate and d is the mean delay (or other measure of performance)
→ Introduce to the objective function importance weights with respect the mean passenger and varies according to the signal timings.
occupancy or the efficiency of passenger occupancy per unit of road space (typically When calculating D, it considers the same importance to each vehicle at the junction.
measured in occupants per Passenger Car Unit (PCU)) for different classes of vehicles Updated version: X
Vehicle Type PCU Value D= wqd (3)
Motorcycle 0.4 where w is calculated according to the mean passenger occupancy of a vehicle:
Passenger Car 1.0 P
Light Goods Vehicle (LGV) 1.0 k q k pk
w= P (4)
Buses and Coaches 2.0 k qk

Sources: TfL (2021). Traffic Modelling Guidelines, Version 4.0, https://content.tfl.gov.uk/traffic-modelling-guidelines.pdf where k is a class of vehicles (e.g. cars and buses), qk is the mean arrival rate of vehicle of
class k and pk is the passenger occupancy of vehicles of class k.
Reduce mean delays (∼ 5 − 8%) in streams with a high proportion of buses.
9 / 33 10 / 33

Signal timings and control (4) Active signal priority

Buses behave differently from other traffic: Priority given in two ways:
▶ Travel at a lower speed
▶ Have to stop at bus stops ▶ Extension
∗ Given if (late) bus arriving during green
Model buses as a different segment of traffic ∗ Extend the current green so that the (late) bus can get through the junction

Capture explicitly the times for stopping,

boarding, alighting, etc ▶ Recall
∗ Given if (late) bus arriving during red
Lead to the calculation of timing plan that ∗ Truncate the current stage to its minimum safe value
considers the trade-off between buses and ∗ Then change to the first stage in which buses have right of way
other traffic

Movements of a platoon of traffic and a bus

11 / 33 12 / 33
Active signal priority (2) Active signal priority (3)
Bus priority may reduce bus journey time, but not necessarily regularity . . .

Extension
▶ Can be carried out without disrupting the stage order relatively small disruptions to
other vehicles
Recall
▶ Have to be carried out by changing the stage order
▶ Could disrupt the progression of other vehicles
▶ Inhibition – rule used to prevent ‘recall’ being granted in two consecutive cycles: Prevent
buses from bunching
▶ Reallocate green times: Compensate loss of green in non-priority stages

Trajectories of buses (Route 29, Central London)

13 / 33 14 / 33

Bus headway control Section 2

Headway based
▶ Forward: looking at the headway from the preceding bus
▶ Backward: looking at the headway from the preceding bus
▶ Combination of two Car following model for CAVs

▶ Headway control is a trade-off between efficiency and regularity (reliability)

▶ With holding strategies, it implies we will need an extra time buffer inserted in the
time-table
∗ Reduce service speed
∗ Increase bus service headway

15 / 33 16 / 33
Car following model for CAVs Car following model for CAVs (2)
▶ Intelligent Driver Model (IDM): a model of desired measures type that calculates the
acceleration based on the current headway or speed against the desired value
▶ Adaptive Cruise Control (ACC): maintains safe following distance based on a predefined
▶ The general formulation of the IDM car-following model is:
speed and gap distance chosen by a drive
" 2 #
vn (t) δn
  
∆X ∗ (t) ▶ Cooperative Adaptive Cruise Control (CACC): The concept of CACC goes beyond
an (t) = amax,n · 1 − −
Vmax,n ∆xn (t) − ln−1 merely using a vehicle’s cruise control to ensure a safe gap from the car ahead by
(5) reducing speed when necessary (ACC). It also involves enabling vehicles to collaborate by
vn (t) · ∆vn (t)
∆X ∗ (t) = s0,n + Tn · vn (t) − p communicating with each other while operating in adaptive cruise control mode.
2 · amax,n · bn
where xn , vn and an are the position, velocity and acceleration of vehicle n; ∆xn and
∆vn are the inter-vehicle spacing and the relative velocity from the leading vehicle; ln−1
is the length of the leading vehicle; ∆X ∗ (t) is the desired minimum gap. Six properties
are inherent to the motion of vehicle n, consisting of the maximum acceleration amax,n ,
the desired velocity Vmax,n , the acceleration exponent δn , the minimum jam
(bumper-to-bumper) distance s0,n , the safety time headway Tn , and the comfortable
deceleration bn .
∗ The Gipps’ model is a safe distance model that determines the velocity or acceleration by
selecting a safe following distance to avoid collision with the lead vehicle
17 / 33 18 / 33

Section 3 Introduction

Capacity management in mixed flow

Notation table is available in the appendix

19 / 33 20 / 33
Introduction (2) Right of Way (RoW) Reallocation - Numerical analysis of theoretical model

21 / 33 22 / 33

RoW Reallocation Case study SUMO simulation results

SUMO simulation

23 / 33 24 / 33
Summary AV size vs road and passenger capacity

▶ A theoretical model based on driving safety constraints is used to calculate the maximum
capacity of the heterogeneous saturated flow. Background
▶ Numerical analyses demonstrate that theoretical capacity increases convexly with AV ▶ AVs are not necessarily the same size as 4-seat passenger cars
penetration rates, which has been proved in theory and cross-validated with SUMO ▶ AVs with reduced sizes → better ensure travel safety
simulation case studies
▶ Debate: some researchers believe that AVs ‘should start small, go slow’
▶ Road capacity can be significantly increased at low or medium penetration rates with
▶ By occupying sub-lanes of sections of roads, smaller AVs would spare more space and
appropriated RoW reallocation strategies.
result in a larger road capacity
▶ Crucial points are identified and impacts of AVs behaviour changes are demonstrated
with sensitivity analysis.

25 / 33 26 / 33

Simulation Simulation (2)

A pure AV case in simulation: different ratios of Normal-sized AVs and Small-Sized AVs Mixed flows of HDVs + AVs (Normal-Sized + Small-Sized AVs) (demand = (2000, 4000))

▶ Small-sized AVs → improve the maximum road capacity during high demand ▶ Low demand
▶ Passengers in AVs (demand = (2000, 4000)) ▶ Medium-level demand: the growth of penetration rates of AVs increases the road
▶ For a full loading of passengers, more large vehicles → the increase of passenger capacity gradually until the vehicle number stabilises.
capacities. ▶ Fully saturated
27 / 33 28 / 33
Section 4 What makes ML appropriate for transportation problems

Nature of transport data:

▶ quantitative and qualitative
Transport systems behaviour:
▶ Interactions among the different components are not fully understood
Machine learning applications ▶ Uncertainties from the human component of the system
Optimisation problems: challenging to solve using traditional mathematical programming
(MP) techniques (e.g. Traffic assignment)
▶ Relationships are hard to specify analytically
▶ Size of the problem and its computational intractability
Finally, the complex nature of transportation systems and the fact that their behaviour
emerges as a result of interactions among the system components makes agent-based
modelling (ABM) techniques quite appropriate for studying the behaviour of the system.

29 / 33 30 / 33

Examples Appendix - Notation table

31 / 33 32 / 33
References

▶ Daganzo, C.F. A headway-based approach to eliminate bus bunching: Systematic analysis and comparisons. Transp. Res. Part B Methodol. 2009, 43,
913–921.
▶ Fancher, P., Ervin, R., Sayer, J., Hagan, M., Bogard, S., Bareket, Z., Hangen, J. & Mefford, M. (1998) Intelligent Cruise Control Field Opera-tional
Test.Final Report. University of Michigan Transportation Research Institute, Ann Arbor.
▶ Jacobson, J. & Sheffi, Y. (1981) Analytical model of traffic delays under bus signal preemption: Theory and application, Transportation Research Part
B: Methodological, 15(2), 127-138.
▶ Li, T., Guo, F., Krishnan, R., Sivakumar, A. & Polak, J. (2020) Right-of-way reallocation for mixed flow of autonomous vehicles and human driven
vehicles, Transportation Research Part C: Emerging Technologies, 115, 102630.
▶ Treiber, M., Hennecke, A. & Helbing, D., (2000) Congested traffic states in empirical observations and microscopic simulations Phys. Rev. E, 62,
1805-1824.
▶ Zhai, X., Guo, F. & Krishnan, R. (2023) An Online Optimal Bus Signal Priority Strategy to Equalise Headway in Real-Time, Information, 14(2), 101.

33 / 33