Measure agent performance

[benderpremier]

The book defines in chapter 2 the notion of PEAS (Performance, Environment, Actuators and Sensors). We are currently missing a way to capture the Performance of our agents.

I am not sure what would be the best way to implement it. The performance is both a function of the Environment state and the Agent state. E.g. in the vacuum world, the book states that we can decide to measure the performance by the number of clean squares at time t, 2 being the best possible score at any point in time. Another possibility would be to take into account the state of the agent. As the agent needs energy to move and clean, the battery level could also be taken into account when measuring the performance.

Here's one way we could do it:

type PerformanceMeasure = (Environment, Agent) => Double

trait Simulation[ENVIRONMENT, PERCEPT, ACTION] {

val environment: Environment [...]

val performanceMeasure: PerformanceMeasure

val performances: Map[Agent, Double]

}

Maybe we should keep as well a history of the agent performance over time. Add the notion of step in time (i.e. one invocation of the run function of the agent)

[benderpremier]

I've created a naive implementation but I am not happy about it for the time being. benderpremier#2 .