Towards Fair and Efficient Traffic Flow Coordination Mechanisms for - - PowerPoint PPT Presentation

▶
towards fair and efficient traffic flow coordination
SMART_READER_LITE
LIVE PREVIEW

Towards Fair and Efficient Traffic Flow Coordination Mechanisms for - - PowerPoint PPT Presentation

Jan 26, 2024 201 likes •481 views

Towards Fair and Efficient Traffic Flow Coordination Mechanisms for 2+1 Roadways M. Aschermann 1 B. Friedrich 2 uller 1 J. P. M 1 Technische Universit at Clausthal 2 Technische Universit at Braunschweig EWGT 2017, Budapest, 4-6.

slide-1
SLIDE 1

Towards Fair and Efficient Traffic Flow Coordination Mechanisms for 2+1 Roadways

  • M. Aschermann 1
  • B. Friedrich 2
  • J. P. M¨

uller 1

1Technische Universit¨

at Clausthal

2Technische Universit¨

at Braunschweig EWGT 2017, Budapest, 4-6. September 2017

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 1 / 27

slide-2
SLIDE 2

Motivation – 2+1 Roadways

Alternating lane segments with one or two lanes per direction – Increase safety of overtaking manoeuvres – Compromise: +40% capacity with one

additional lane

2+1 systems mandatory in Germany for newly constructed urban roadways Existing roadways to be extended during normal maintenance phases

[Arbeitsgruppe Straßenentwurf, 2013, BASt, 2013]

B 54 near Steinfurt

(picture: public domain)

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 2 / 27

slide-3
SLIDE 3

Motivation – Managed Lanes

Example: Managed HOT lane on Interstate 15 with variable pricing

(picture: Chevy111, CC BY-SA 4.0, cropped)

High-occupancy vehicle (HOV)/high-occupancy toll (HOT) lanes Access based on dynamic rules, e.g. fixed or dynamic congestion pricing

[de Palma and Lindsey, 2011, Rouhani, 2016]

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 3 / 27

slide-4
SLIDE 4

Assumptions – Autonomous Vehicles and Smart Infrastructure

We envision a (near) future scenario comprising Autonomous vehicles – Automated Driving Systems (ADS) commonly available – Classified as level 3 and above [SAE International, 2016]

(⇒ autonomous overtaking and lane changing manoeuvres)

– Drivers provide individual preferences, e.g. – desired speed – acceptable time loss Communication infrastructure on HOV/HOT lanes – 4G and upcoming 5G/G5 networking technology – Real-time traffic observation

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 4 / 27

slide-5
SLIDE 5

System vs. User Level Optimisation

System Level: Traffic Management Goals: Optimise efficiency on 2+1 roadways and HOT/HOV lanes – minimise travel time losses, maximise traffic flow, reduce congestions Mechanisms: Change rules dependent on situations – e.g. denying access to overtaking lane for slow vehicles. But: Consider fairness! User Level: Drivers Minimise dissatisfaction, quantifiable by travel time loss – but individual preferences (time pressure / relaxed driving) also relevant Acceptable time loss depending on time pressure

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 5 / 27

slide-6
SLIDE 6

Research Goals

Identify potentials of optimising the 2+1 system by means of coordination – Minimisation of unfairness, inefficiency and dissatisfaction Considered input-orderings of vehicles by desired speeds of drivers

  • 1. best-case: already optimal ordering of vehicles
  • 2. random-case: arbitrary ordering of vehicles
  • 3. worst-case: ascending ordering of vehicles

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 6 / 27

slide-7
SLIDE 7

State of the Art and Related Work

2+1 Roadways – [Irzik, 2010a, Irzik, 2010b] Coordination and management of vehicles on a microscopic level – Autonomous Intersection Management (AIM) platform

[Dresner and Stone, 2004, Dresner and Stone, 2005, Dresner and Stone, 2008]

– Policies for linked intersections [Hausknecht et al., 2011a] – Extension Semi-AIM: More efficient intersection management [Au et al., 2015] – High level of compliance to policies on managed lanes (MLs) with only

painted barriers, i.e. obeying the rules implemented by a coordination service entity (CSE) [Halvorson and Buckeye, 2006]

– Coordination mechanism for handling varying demand (e.g. commuter

traffic) by employing a dynamic lane reversal for maximising throughput

[Hausknecht et al., 2011b]

Driver preferences – [Ringhand and Vollrath, 2017] investigated driver preferences for route choices

and analysed acceptance threshold regarding travel time loss

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 7 / 27

slide-8
SLIDE 8

Design Decisions – Coordination Service Entity (CSE)

Vision: CSEs, governing 2+1 roadway segments, for receiving access requests granting access to the overtaking lane based on rules

  • bserving traffic flow, driver satisfaction, coordination efficiency and

fairness

30 80 100 40 40 30 100 80

Coordination Service Entity (CSE)

vehicles: send request for clearance, {preferences, capabilities} 40 30 100 80 vehicles: report satisfaction CSE: optimised rules system & user level maintain fairness

Phase 1: Registration Phase 2: Optimisation Phase 3: Assessment

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 8 / 27

slide-9
SLIDE 9

Model – Basic Definitions

  • 1. Inefficiency
  • 2. Dissatisfaction
  • 3. Unfairness

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 9 / 27

slide-10
SLIDE 10

Model – Inefficiency

Inefficiency derived from relative time loss and scenario parameters relative time loss scenario parameters

  • ⇒
  • Vehicles

⇒

250 500 750 1000 500 1000 1500 2000 demand inefficiency

colour 2

Relative time loss: normalised by optimal travel time ⇒ comparability

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 10 / 27

slide-11
SLIDE 11

Model – Dissatisfaction I

We designed our dissatisfaction model related to [Ringhand and Vollrath, 2017] – sigmoid function and threshold when satisfaction turns into dissatisfaction Dissatisfaction derived from relative time loss and time loss threshold relative time loss △ time loss threshold ♣

  • ptimal travel time ♠

smoothing factor ρ        ⇒ sigmoid function: 1 1 + e(−△+♣·♠)·ρ ⇒

0.00 0.25 0.50 0.75 1.00 80 90 100 110 120 relative travel time loss in percent dissatisfaction

1 colour

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 11 / 27

slide-12
SLIDE 12

Model – Dissatisfaction II

Dissatisfaction parametrisation

0.00 0.25 0.50 0.75 1.00 30 60 90 120

relative travel time loss in percent dissatisfaction

vehicle type

passenger tractor truck

Time loss thresholds – passenger vehicles: 0.2

[Ringhand and Vollrath, 2017]

– trucks: 0.1 – tractors: 1.0 Optimal travel time:

length(r) v max(a)

Relative time loss:

TT act(a,r)−TT opt(a,r) TT opt(a,r)

Smoothing: ρ = 0.5

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 12 / 27

slide-13
SLIDE 13

Model – Unfairness

Interquartile range (IQR)1 as a general indicator of unfairness in the system Unfairness derived from inter-quartile distance of relative time losses relative time- losses of vehicles ⇒ IQR (H-Spread1) ⇒ 26 40 54 34 36 41 46 36 42

  • unfairness = |36 − 42| = 6

Domain independent Can be applied independently from the underlying system/scenario ⇒ Robustness against unfair configurations IQR ≈ 0: Fair system, i.e. no particularly (dis)advantaged drivers IQR > 0: Certain drivers are (dis)advantaged

1H-Spread see [Weisstein, 2017] Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 13 / 27

slide-14
SLIDE 14

Research Questions I

Hypotheses H.1 Best-case ordering yields close to optimal results. H.2 Random- and worst-case ordering negatively affects optimisation dimensions with rising demand.

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 14 / 27

slide-15
SLIDE 15

Research Questions II

Research Questions RQ.1 Can optimisation potentials from traffic management and drivers’ perspectives be identified, i.e. at what traffic service levels it is sensible to apply optimisations? RQ.2 Can an estimation be given on how much improvement could theoretically be achieved? RQ.3 Does the ordering of vehicles by their maximum driving speed play an important role regarding room for optimisation?

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 15 / 27

slide-16
SLIDE 16

Evaluation – Experimental Study Setup

B 210: 2+1 roadway with parameters as in [Irzik, 2010b, p. 167] length switches speed limit vehicles/lane/hour 6800 m 4 100 km/h 514.¯ 6 Consistent driver model: sigmoid with 20% threshold for passengers

[Ringhand and Vollrath, 2017]

Vehicle type distribution: – 80% passenger vehicles – 15% trucks – 5% tractors Tested demand of 200 to 2000 vehicles/lane/hour in steps of 150 – covering service level classification A to E [TDM Encyclopedia, 2017]

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 16 / 27

slide-17
SLIDE 17

Evaluation – Simulation of 2+1 Roadways with SUMO

  • 1. Modelled scenario with Simulation of Urban MObility (SUMO)
  • 2. Implemented our models and a runtime with Python to

– Receive vehicle information – Compute current dissatisfaction, fairness and efficiency – Write/update vehicle information

  • 3. Connected our runtime with SUMO via Traffic Control Interface (TraCI)

CoLMTO Execution Model

Input

Configuration and data model generator

YAML YAML YAML SUMO routes XML

Execution

SUMO Coordination Service Entity SUMO Coordination Service Entity

Coordination Service Entity Runtime SUMO

Output

Statistics

Time loss Dissatisfaction Unfairness HDF5 YAML Input of scenario specification and simulation preferences Experiment config Scenario config Vehicle types SUMO network Active rules XML

Cooperative Lane Management and Traffic flow Optimisation (CoLMTO)

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 17 / 27

slide-18
SLIDE 18

Discussion of Results – Hypothesis H.1

Passenger vehicles are the main beneficiary in

  • ur scenarios and comprise the major share

(80%) ⇒ focus of discussion

Hypothesis (H.1) Best-case ordering yields close to optimal re- sults.

⇒ Potentials are slim and not worthy of further consideration, supported by results for the best case ordering of vehicles.

  • best case ordering

relative time loss best case ordering dissatisfaction best case ordering inefficiency best case ordering unfairness

500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 0.00 0.25 0.50 0.75 1.00 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 2.5 0.00 0.25 0.50 0.75 1.00 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 18 / 27

slide-19
SLIDE 19

Discussion of Results – Hypothesis H.2

Hypothesis (H.2) Random- and worst-case order- ing of vehicles negatively af- fects optimisation dimensions with rising demand. ⇒ supported by results. Observations Peaks with declining slopes Decline after peak (peak•):

– attributed to vehicles

avoiding overtaking lane

⇒ less friction induced by

lane changes Local minimum after peak: min(peak•)

  • worst case ordering

relative time loss worst case ordering dissatisfaction worst case ordering inefficiency worst case ordering unfairness random ordering relative time loss random ordering dissatisfaction random ordering inefficiency random ordering unfairness

500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 2.5 0.00 0.25 0.50 0.75 1.00 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 2.5 0.00 0.25 0.50 0.75 1.00 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 vehicles/lane/hour Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 19 / 27

slide-20
SLIDE 20

Discussion of Results – Research Question RQ.1

Research Question RQ.1 Can

  • ptimisation

potentials from traffic management and drivers’ perspectives be identi- fied, i.e. at what traffic service levels it is sensible to apply op- timisations?

Potentials to reduce relative time loss and unfairness ⇒ indirectly dissatisfaction

and inefficiency

Optimise where X(demand) ≥ min(peak•)

  • worst case ordering

relative time loss worst case ordering dissatisfaction worst case ordering inefficiency worst case ordering unfairness random ordering relative time loss random ordering dissatisfaction random ordering inefficiency random ordering unfairness

500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 2.5 0.00 0.25 0.50 0.75 1.00 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 2.5 0.00 0.25 0.50 0.75 1.00 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 vehicles/lane/hour Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 20 / 27

slide-21
SLIDE 21

Discussion of Results – Research Question RQ.2

Research Question RQ.2 Can an estimation be given on how much improvement could theoretically be achieved?

Estimated improvement ≈ |peak − min(peak•)|

  • worst case ordering

relative time loss worst case ordering dissatisfaction worst case ordering inefficiency worst case ordering unfairness random ordering relative time loss random ordering dissatisfaction random ordering inefficiency random ordering unfairness

500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 2.5 0.00 0.25 0.50 0.75 1.00 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 2.5 0.00 0.25 0.50 0.75 1.00 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 vehicles/lane/hour Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 21 / 27

slide-22
SLIDE 22

Discussion of Results – Research Question RQ.3

Research Question RQ.3 Does the ordering of vehicles by their maximum driving speed play an important role regard- ing room for optimisation?

Worst case ordering more sensitive to losses due to lane-change friction – shows distinct

min(peak•) compared to random case

  • worst case ordering

relative time loss worst case ordering dissatisfaction worst case ordering inefficiency worst case ordering unfairness random ordering relative time loss random ordering dissatisfaction random ordering inefficiency random ordering unfairness

500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 2.5 0.00 0.25 0.50 0.75 1.00 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 0.0 0.5 1.0 1.5 2.0 2.5 0.00 0.25 0.50 0.75 1.00 500 1000 1500 2000 0.0 0.5 1.0 1.5 2.0 vehicles/lane/hour Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 22 / 27

slide-23
SLIDE 23

Conclusion and Outlook

We modelled the combination of driver preferences and optimisation goals

  • f managed lanes

We conducted a pre-study on simulated 2+1 manoeuvres by using our own framework with SUMO to estimate optimisation potentials of coordination We identified potentials to reduce driver dissatisfaction while maintaining fairness – policy-based fine-tuning is necessary to avoid imbalances between

  • ptimisation goals

Future Work Baseline for further studies to enhance the effectiveness of chosen rules Integrate a fine-grained control for accessing managed lanes based on policies attractive to drivers and effective coordination mechanisms for a cooperative traffic management

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 23 / 27

slide-24
SLIDE 24

Thank You For Your Attention

30 80 100 40 40 30 100 80

Coordination Service Entity (CSE)

vehicles: send request for clearance, {preferences, capabilities} 40 30 100 80 vehicles: report satisfaction CSE: optimised rules system & user level maintain fairness

Phase 1: Registration Phase 2: Optimisation Phase 3: Assessment

Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 24 / 27

slide-25
SLIDE 25

References I

Arbeitsgruppe Straßenentwurf (2013). Richtlinie f¨ ur die Anlage von Landstraßen (RAL). FGSV Verlag GmbH. Au, T.-C., Zhang, S., and Stone, P. (2015). Autonomous intersection management for semi-autonomous vehicles. Handbook of Transportation. Routledge, Taylor & Francis Group. BASt (2013). Neue Richtlinien f¨ ur die Anlage von Landstraßen vorgestellt. Accessed: 2017-03-20. Chevy111 (2014). Express lanes along Interstate 15 southbound near Escondido, CA. Accessed: 2017-08-30, By Chevy111 (Own work) [CC BY-SA 4.0 (http://creativecommons.org/licenses/by-sa/4.0)], via Wikimedia Commons. de Palma, A. and Lindsey, R. (2011). Traffic congestion pricing methodologies and technologies. Transportation Research Part C: Emerging Technologies, 19(6):1377–1399. Dresner, K. and Stone, P. (2004). Multiagent traffic management: A reservation-based intersection control mechanism. In Proceedings of the Third International Joint Conference on Autonomous Agents and Multiagent Systems-Volume 2, pages 530–537. IEEE Computer Society. Dresner, K. and Stone, P. (2005). Multiagent traffic management: An improved intersection control mechanism. In Proceedings of the fourth international joint conference on Autonomous agents and multiagent systems, pages 471–477. ACM. Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 25 / 27

slide-26
SLIDE 26

References II

Dresner, K. and Stone, P. (2008). A multiagent approach to autonomous intersection management. Journal of artificial intelligence research, 31:591–656. Halvorson, R. and Buckeye, K. R. (2006). High-occupancy toll lane innovations: I-394 mnpass. Public Works Management & Policy, 10(3):242–255. Hausknecht, M., Au, T.-C., and Stone, P. (2011a). Autonomous intersection management: Multi-intersection optimization. In 2011 IEEE/RSJ International Conference on Intelligent Robots and Systems, pages 4581–4586. IEEE. Hausknecht, M., Au, T.-C., Stone, P., Fajardo, D., and Waller, T. (2011b). Dynamic lane reversal in traffic management. In 2011 14th International IEEE Conference on Intelligent Transportation Systems (ITSC), pages 1929–1934. IEEE. Irzik, M. (2010a). Layout of 2+1-routes in Germany – New findings. In 4th International Symposium on Highway Geometric Design, Valencia, Spain, pages 2–5. Irzik, M. (2010b). ¨ Uberholverhalten auf 2+1-Strecken: Ein Beitrag zur Gestaltung von dreistreifigen Landstraßen. Schriftenreihe des Instituts f¨ ur Verkehr und Stadtbauwesen, TU Braunschweig, IV+145S(55). Ringhand, M. and Vollrath, M. (2017). Investigating urban route choice as a conflict between waiting at traffic lights and additional travel time. Transportation Research Procedia, 25:2432–2444. World Conference on Transport Research - WCTR 2016 Shanghai. 10-15 July 2016. Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 26 / 27

slide-27
SLIDE 27

References III

Rouhani, O. M. (2016). Next generations of road pricing: Social welfare enhancing. Sustainability, 8(3):265. SAE International (2016). Taxonomy and definitions for terms related to driving automation systems for on-road motor vehicles. Accessed: 2017-03-20. TDM Encyclopedia (2017). Congestion reduction strategies. Accessed: 2017-03-20. Weisstein, E. W. (2017). H-spread. From MathWorld – A Wolfram Web Resource. Woehlecke (2009). B 54 bei Steinfurt. Accessed: 2017-08-30, By Woehlecke at German Wikipedia (Own work (Original text: selfmade)) [Public domain], via Wikimedia Commons. Aschermann, Friedrich, M¨ uller | TU Clausthal, TU Braunschweig | Traffic Flow Coordination Mechanisms 27 / 27