LLNL / Abmarl

Agent Based Modeling and Reinforcement Learning
Other
61 stars 18 forks source link

Hoteling's Law Example #472

Open rusu24edward opened 11 months ago

rusu24edward commented 11 months ago

In Epic #2

Unknown-3

Hotelling’s Law is a competitive game where sellers “sell goods” in a grid. Each seller sets the price of its goods, and all the nearby cells are motivated to buy from it with two principles: distance from the seller and its price. Each seller is motivated to set its price such that it will make a profit and attract the largest customer base. It can also move around the grid to optimize its location at a moving cost. The multiple sellers in the grid must compete for customer domination. A seller goes out of business if it runs out of money. The agent controls a single seller, and the other sellers have scripted behavior.

Need to be able to change agents' encodings Blocked by #367 Blocked by #464

rusu24edward commented 7 months ago

Grid full of buyer agents Seller agent randomly placed Sellers can overlap buyers, but not each other Seller has "income", which is correlated to how many buyers it has in each step; income added to money Each step costs some amount of money Seller can move, which costs lots of money If seller runs out of money, then it goes out of business Sellers can set price, and the sim updates the buyers automatically (based on distance and price). Each consumer adds up the price and distance from each store, and then chooses to go to the store that offers the lowest sum. Ties broken randomly. Sellers observe the grid and their own money and price that they have set. Sellers can observe each others prices? (maybe they can take a "gain insight" act that costs money and reveals the other agents' price for a few steps)? Buyers have spending limit: if there is no store that meets its requirements, then it will just not buy anything.