Raking the Agents' Solution

The Quality of a Solution

In the first version of the protocol, the quality Q(s) of an order price solution s can be defined by:

$$Q(s) = f_{\texttt{surplus}}(s) + f_{\texttt{speed}}(s)$$

The agent's goal is to maximize a utility function defined as:

$$\Omega = \max \left({Q(s)}\right) - \min(f_{\texttt{costs}}(s))$$

At the conclusion of the auction, solutions are ranked in descending order, and the agent with the highest score (i.e., the largest utility function) wins the round.

In future versions of the protocol, the quality of a solution will include more features.

Based on this function, the agent's reward will then be computed as a multidimensional agent reward.

Additionally, each solution's score will contribute to the agent's overall reputation score in the Arena.