Large-scale collaborative vehicle routing.

Area of Science:

• Operations Research
• Artificial Intelligence
• Logistics and Supply Chain Management

Background:

• Carrier collaboration, particularly via platforms, offers substantial reductions in transportation costs and emissions.
• Previous investigations into these gains were limited to small-scale scenarios with few carriers.

Purpose of the Study:

• To develop and evaluate auction-based methods for large-scale dynamic collaborative pickup and delivery problems.
• To assess the benefits and strategic behavior implications of widespread carrier cooperation.

Main Methods:

• Utilized multi-agent systems and combinatorial auction techniques.
• Employed iterative single-order auctions and compared them with central combinatorial auctions.
• Integrated local and central auction approaches with small bundle auctions for enhanced solution quality.
• Evaluated performance using a real-world dataset of over 12,000 orders.

Main Results:

• Achieved up to 77% reduction in travel costs with 1000 collaborating carriers, significantly exceeding prior small-scale findings.
• The proposed multi-agent system with local auctions outperformed central combinatorial auctions for large-scale instances.
• Bundling strategies in decentralized systems yielded additional travel cost gains of up to 13% for 1000 carriers.
• Identified potential for strategic behavior (cheating) but also proposed mitigation strategies.

Conclusions:

• This research provides the first assessment of large-scale carrier cooperation benefits and proposes a viable approach.
• Auction-based methods, especially when combined and adapted for scale, are effective in maximizing collaborative logistics gains.
• While strategic behavior is a concern, it can be managed, ensuring the integrity of cooperative systems.