Stopping Criteria Freight Transportation Network Network Data Network Simulation Models Network states Optimization: Minimum cost Route Load Balancing Using a Co-Simulation/Optimization/Control Approach Petros Ioannou University of Southern California Los Angeles, CA 90089 Email: ioannou@usc.edu
Freight Transportation System RAIL OCEAN ROAD PORT
Motivation Freight transport is fundamental to human survival especially in Metropolitan areas Globalization let to a rapid increase in import and export activities Freight is transported via rail, trucks, ships and air (high value small size) Freight rail and trucks share the same networks as passengers Trucks are big, have slower dynamics, disturb traffic during merging and changing lanes. They pollute more and waste more fuel. Current road network treats trucks more or less like every other vehicle (exception are some rules in some places) Congestion and safety are concerns where truck volumes are high
Current Issues Both rail and road networks are in need of extra capacity especially at Metropolitan areas Both networks exhibit low peak and high peak traffic which has temporal but sometimes spacial characteristics too This indicates load imbalances across the networks that are due to several reasons. Some reasons are: Lack of information and lack of a system that can help identify available free capacities and balance loads across the network in a more efficient way
Trends Intelligent Transportation Technologies Information Systems and Internet GPS, FRIDs, Bluetooth, communications Connected vehicles Traffic Management Systems Vehicle Technologies: ADAS, Vehicle automation. New sensors and software tools are coming up and maturing in accuracy and reliability
Problem Statement The current Freight transportation network is unbalanced w.r.t temporal and spacial coordinates Q: Can upcoming ITS technologies help? ANS: Yes provided the appropriate intelligence and tools are developed
Technical Challenges Both rail and road networks involve complex operations They are complex dynamic systems with time varying characteristics Lack of adequate sensors and data (changing with ITS) Traditional Approach to handle the problem Simple models and adhoc techniques What happened yesterday or last week will be repeated
Approach for control and optimization of complex systems The use of simple mathematical models is a fundamental practice in every engineering design for controlling and optimizing systems The requirement is that these models are simple enough to understand and lead to less complex designs that are easy to implement yet they are complex enough to represent the dominant characteristics of the system When it comes to complex networks such as the transportation network simple models cannot always capture the dynamics, interconnections and complexity involved
Approach for control and optimization of complex systems The availability of fast computers and software tools point to the next step in the evolution of models from very simple during the Analog years to more Complex during the Digital era to: Simulation models as part of the control and optimization feedback system To handle bigger complexity and complex dynamics and interconnections with the limit the computational speed relative to the bandwidth of the control system.
Approach for control and optimization of complex systems Controller Stopping Criteria Real Network System Network Data Network Simulation Models Network states Optimization
Approach for control and optimization of complex systems Final Decision Load balancing Controller Stopping Criteria Freight Transportation Network Network Data Network Simulation Models Network states Optimization: Minimum cost Route
Approach for control and optimization of complex systems Issues Validation of simulation models. Scalability and complexity Balancing control design Optimization Stability and robustness of closed loop system Convergence to equilibrium Speed of convergence
Approach for control and optimization of complex systems F Optimization part: Service network Origin N o S 1 S2 S 3 N 1 N 2 S 5 S 6 N 4 S 8 S 9 S 10 S 11 Destination N D S 4 N 3 S 7 N 5 S 11
Approach for control and optimization of complex systems Optimization part: Find minimum cost route for a given load over a time interval T in the future. The cost depends on the states of the networks i.e link flows in a complex fashion that cannot be represented by a simple model. The simulation model estimates/predicts the states of the network at each iteration The optimization of minimum cost route is solved pointwise wrt each iteration.
Approach for control and optimization of complex systems Load Balancing Control (LBC): Currently adhoc LBC: Starts with an initial load distribution. SM: estimates the states of the network Opt: computes the minimum cost route LBC: Passes loads to the minimum cost route to equalize cost with previous minimum cost route(s)
Approach for control and optimization of complex systems Stopping Criteria The balancing approach guarantees convergence as the cost function is bounded from below and is non increasing due to the procedure. In practice there may be limitations and the iterations may have to stop after a certain time or after additional improvements are small.
Train Station PHYSICAL D4 D5 D6 D1 D2 D3 Train Station C B A SC SC SC D A T A CYBER Coordinator Load Balancing D A T A Example
EXAMPLE Origin: 3 terminals A (1020), B(1020), C(1020) Destinations: D1 (350), D2(450), D3(400), D4(600),D5(700), D6(560) Shippers: Three, one for each terminal Routes: Road Network and Partial Rail (5 trains with capacity of 50 containers each) Network Complexity: 12824 links and 4747 nodes Simulation Model: Macroscopic based on VISUM. Use historical data and dynamic assignment to tune it. Cost: Minimize total $ cost based on real data which depends on travel times and mode
EXAMPLE.
EXAMPLE Uncoordinated Coordinated.
Conclusions New approach for controlling and optimizing systems where simple mathematical models are replaced with simulation models in a closed loop configuration Application to Freight Load Balancing problem Todays available technologies and computational techniques support the potential of the method for its widespread use Already successfully applied traffic light control of the city of Chania, Crete, Greece (currently operating) Energy building in Germany as part of an EU project
Stopping Criteria Freight Transportation Network Network Data Network Simulation Models Network states Optimization: Minimum cost Route Thank You