|
Domain |
Explanation |
|
Speaker |
- Dr Lee Der Horng, assistant professor, Department of Civil Engineering, NUS, specializes in microscopic traffic simulation, applying various classification, clustering, heuristics, artificial intelligence and other data mining methodologies to analyse dynamic, online traffic data flows. Dr Lee is presently using his innovative tools to tackle critical issues in urban traffic management which would ensure a smoother ride for commuters traveling to and from work. His influential and international standing in microscopic traffic simulation and ITS has made him well sought after by transportation authorities worldwide. Dr. Lee and his ITVS (Intelligent Transportation and Vehicles Systems) teammates at NUS are discussing and working with local and foreign research and development agencies, universities as well as companies in car manufacturing, IT (information technology), software, and transportation to further materialize tangible benefits of ITS in our day to day living.
- Dr Lee Der-Horng has been chosen as one of the world’s 100 Top Young Innovators by Technology Review, MIT’s Magazine of Innovation. The TR100, chosen annually by Technology Review, MIT’s award-winning magazine of innovation, consists of 100 young individuals whose innovative work in business and technology has a profound impact on today’s world. Dr Lee is an associate editor of IEEE Transactions on Intelligent Transportation Systems and the editor of Urban and Regional Transportation Modeling published by Edward Elgar.
|
|
Abstract |
- Parallel traffic simulation is an application of parallel computing techniques that aims to decrease the computation time by engaging different processors of a multiprocessor system or different computers of a network. Very few traffic simulation models have this capability of parallel simulation. However, it is possible to upgrade a simulation program that is not capable of running parallel simulation by applying the method proposed in this paper. The method involves dividing the network into regions and simulating each region under a separate instance of the program. Suitable inter-process communication techniques are employed to exchange data between different regions and synchronize time among the different regions. A significant increase in simulation speed is seen when the proposed method is applied to Paramics, a microscopic time-stepping simulation program.
|
|
Contents |
- Foreword
- Simulation
- Issues
- Traffic simulation: classification, attributes & limitations
- ITVS
- Parallel simulations with multiple instances of program simulation
|
|
Foreword |
- When to simulate:
- Analytical model is too large-scale & complex: too many interactions
- Too computationally-intensive
- Balance of maths rigour & model flexibility
- Maths rigour: e.g. user equilibrium à criteria & governing equation à Beckmann transformation solution à properties well-known
- Model flexibility: changed & modified rapidly to show how variable affects traveler choice à more difficult to be into single traffic model
- Resource complexity: computers (available for micro-simulation: more details, more data, constraint practicality into traffic modeling); more interactions; more data; more financial
|
|
Simulations |
- Aims:
- To mimic component operations
- System interactions
- Driven by random nos. & variates
- Only simulates, not replicates what would actually happen
- Only approximates with stochastic process
- Not general, but very specific
- Event-based: decision-making triggered by events happening, where attributes are only refreshed or triggered by certain events
- Time-based: attributes are refreshed in time-steps, irregardless of events
- More for microscopic simulations: time step ¬0.5s
- Less time interval, more computations
|
|
Issues |
- Verification: logical correctness
- Calibration: to fit model parameters & able to shift between different contexts, time & behaviours
- Validation: O-D tables: narrow down complexity, data à fitness of model prediction with actual data measured
- Variation reduction: similar to model condensation
- Stochastic approximation due to randomness
- Aim for less variation for essential characteristics from system flows, technical reduction & condensation (no. & lengths of runs)
- System warm-up time: from empty network to full system
- Depending on: experiences, network structure
- Sample runs first to pick out essential characteristics
|
|
Traffic simulation |
- Alternative & complementary to analytical modeling, depending on structuring
- More suitable for complex systems
- Deliver integrated & comprehensive traffic analysis & evaluation
- Classification:
- Macroscopic: low fidelity, less details, essential characteristics, qualitative; how system performs without individual focus; link travel pattern unchanged & changes at intersections
- Microscopic: high fidelity, much details, many individual characteristics & focus; changes at link travel pattern & intersections
- Mesoscopic: mixed between the two
- Attributes:
- Car-following: by mean headway, reaction times
- Lane-changing
- Route choice
- Travel demand
- Mechanical
- Traffic control devices
- Limitations:
- Much expensive: more computationally-intensive
- Problem-scale: balance between expectations & resources
- Data availability: customize & analyse
- Output analysis
|
|
ITVS |
- Real-time online traffic simulation system:
- Pre-processing:
- Data à screening à fusion: historical & static traffic à online real-time database à incident simulatio
- Processing:
- Data mining & parameters à expert system à real-time traffic simulation
- Post-processing: results, analysis for decision-making
|
|
Parallel simulation with multiple instances of program simulations |
- Background:
- Parallel simulation uses distributed or parallel computing
- Computer load shared amongst processors &/or computers simultaneously
- Advantages:
- Faster speed
- Large-scale: more realistic
- More real-time possible
- Methods: 3 levels
- Data: multiple data fusion
- Machines: single vs. multiple processors &/or computers
- Programs: capabilities for parallel computing within simulation algorithm
- Programs:
- Approaches:
- Conventional: implemented in Paramics already
- One traffic network, but 3 zones simulated by 3 separate processors
- Direct connection
- Proposed approach:
- 1 network, but 3 separate zones
- Each zone by one processor
- IPC: interprocess communication for networking between the 3 processors
- No shared memory
- No threads
- Purposes:
- Data exchange & time synchronization
- Vehicle exchange at system boundaries: copy vehicle from one zone & re-release at the adjacent zone
- Test case:
- Grid network, signalied intersections, loop detectors
- Sun 4-processors: 3 zone processors, 1 IPC
- Using turn-count based (non O-D) simulation: by traffic count only
- Performance:
- Faster: more processors, less vehicles, not proportional
- Algorithm speed-up more: more vehicles, more processors
- Apply to larg-scale OD-based environment
|