Efficient and Equitable Exchange in Air Traffic Management Plan Repair using Spender-signed Currency

Abstract

Air traffic management (ATM) is concerned with planning of air traffic in the air and on the ground. At any moment up to the moment of execution, a plan may become infeasible, for instance as a result of delays or mechanical failures. In that case a plan needs to be repaired. This may involve delays, advances and resource changes for one or more aircraft. A current trend in aviation research is that of Collaborative Decision Making. In this philosophy, operational decisions are the result of a negotiation between all parties involved, rather than taken by a single authority. We studied the problem of facilitating collaborative plan repair in ATM. Airlines are competitors, who care primarily about their own efficiency. However, they can often help each other. Airlines will be willing to help each other if they know that given help will be reciprocated. In fact, the process of giving and receiving help can be viewed as a social exchange process. An established fact in the social sciences is that a social exchange is perceived to be equitable by its participants if each agent gives as much to others as it receives from others. We derive from this a measure for the equity and thus the acceptability of the exchange. Money is an administrative system for reciprocal exchange. If agents pay each other for received services, and the prices correspond to the utilities of the tasks performed, the financial balances of agents indicate their entitlements and debts to the society. We show that, if balances of agents are bounded from below, the resulting exchange is bounded inequitable. However, agents might choose to lie about utilities. Especially if a service is of much value to the receiver, an agent can increase its price. We call this exploitation. An exploiting agent earns more money than it should, and can spend this to receive extra help from others. The result is that the inequity of exchange can grow arbitrarily large. Agents could in principle penalize exploiters if they can estimate the amounts of exploitation well enough. But this strategy is not robust, as it is attractive to deviate from it. This can result in ever growing inequity. We propose a new currency system, the spender-signed currency system. In this system, each agent signs every credit is uses. The value of a single credit is determined by multiplying the reputations of the agents that have signed that credit. If an exploiter is given a low reputation by the others, it will find that the credits it uses lose value. In this way exploiters are penalized. Importantly, it is not attractive to deviate from the penalizing strategy. We prove the effectiveness of the spender-signed currency system analytically for a relatively specific case, and empirically for a more general case. We show that the spender-signed currency system functions well under reduced levels of trust. Also, we explore some other interesting properties. For instance, the system enables agents to exchange, also indirectly, predominantly with allies. This in contrast with standard currency, where agents usually trade, indirectly, with all other agents, including rivals.