Presentation is loading. Please wait.

Presentation is loading. Please wait.

Cooperative Transport Planning An Experimental Environment Jonne Zutt ( ) TU Delft Parallel Distributed Systems CABS Project.

Similar presentations


Presentation on theme: "Cooperative Transport Planning An Experimental Environment Jonne Zutt ( ) TU Delft Parallel Distributed Systems CABS Project."— Presentation transcript:

1 Cooperative Transport Planning An Experimental Environment Jonne Zutt ( J.Zutt@its.tudelft.nl ) TU Delft Parallel Distributed Systems CABS Project

2 Contents Transport Planning Problem Architecture Algorithm MARS (TNO-TPD) The demo Results (concept)

3 Transport Planning Problem Infrastructure Trucks (capacity, speed) Orders (source, destination, profits, volume, timewindow) Plans (sequence of locations) Costs (distance, nr of trucks, timewindows) TPP Architecture Algorithm MARS Demo Results

4 TPP (2) Objective: Find a plan for every truck that minimizes the total costs, such that all orders are pickup up and delivered. TPP Architecture Algorithm MARS Demo Results

5 Broker Company 1Company 3 Company 2 Architecture Customer TPP Architecture Algorithm MARS Demo Results

6 Planning TPP Architecture Algorithm MARS Demo Results OrdersPlans First Phase Second Phase between all trucks of the company; Insertion, savings, incremental local optimization. cooperation between trucks in a coalition; clustering is used to decide which agents will cooperate; exchanging of orders.

7 Multi-Agent Real-time Simulator TNO TPD Written in the Java language, platform independent Scalable Two parts: (i) Basic (generic) simulator (ii) Experiment TPP Architecture Algorithm MARS Demo Results

8 MARS Experiment 1.Entities 2.Infrastructure 3.Scenario 4.Visual Model TPP Architecture Algorithm MARS Demo Results

9 TPP Architecture Algorithm MARS Demo Results

10 TPP Architecture Algorithm MARS Demo Results 17 runs, 10 random orders, 9 trucks. Average solution quality = 103.5%. Average cooperation gain = 76.1%.

11 Future Work Implement other heuristics for creating the initial plans; Try other forms of exchange, e.g. exchange combinations of orders; Incident handling: trucks are not executing exactly as planned.

12 The End

13 TPP (1) Directed graph G = ( Locations, Arcs ). Orders O = { id, source, target, profits, volume, timewindow }. Trucks T = { id, capacity, speed }. Plan = Cost function = (distance, nr of trucks, timewindows) TPP Architecture Algorithm MARS Demo Results

14 TPP (2) Total costs = The problem is to find a plan for every truck, such that all orders are picked up and delivered, thereby minimizing the total costs (i.e. the sum of the costs of the plans of all individual trucks). TPP Architecture Algorithm MARS Demo Results

15 Multi-Agent Real-time Simulator Basic Simulator: (Un)register objects Communication Loading experiment Initializing – Simulation steps – Termination TPP Architecture Algorithm MARS Demo Results


Download ppt "Cooperative Transport Planning An Experimental Environment Jonne Zutt ( ) TU Delft Parallel Distributed Systems CABS Project."

Similar presentations


Ads by Google