Research Article
Reinforcement Learning in an Environment Synthetically Augmented with Digital Pheromones
Algorithm 2
Agent action selection algorithm.
| Algorithm: agentAct | | Input: agent location, loc | | Returns: destination grid cell, | | S ← get States In Neighborhood(loc) // set of augmenter states in agent’s neighborhood | | ← max() // augmenter signature with highest learned value | | if == null // no distinct maximum state value exists | | = Level-Bias() // augmenter signature with highest augmenter level | | ← get Grid Cells() // set of all neighborhood grid cells having state s | | ← Event-Bias() // grid cell from with highest Event pheromone level | | if == null // all Event pheromone levels are equal | | ← randomGrid() // choose at random from | | ok ← auction(value(), ) // enter auction with state value and destination grid | | // auction returns true if the grid cell is unclaimed or if the | | // grid cell is the last destination option for the agent | | if not(ok) agent Act(loc) | | return () |
|
