-
Notifications
You must be signed in to change notification settings - Fork 0
Framework introduction – Vendors
This page will introduce the various vendors we have.
Copied over from other page, needs reformatting
The marketplace is a stage of competition between different vendors. The vendor component (that could act as a competitor or an agent) assumes the active role of our marketplace that tries to achieve the maximal profit by selling its items to customers. Actually, there is no difference between an agent and a customer. Basically, an agent is a focused vendor whereas competitors operate outside of the viewers focus.
Since it operates in an active manner, we have the possibility to observe our agent while competing against the other vendors by depicting valuable analyze-data such as mean/max/min profit per step, etc. Besides, you are able to display such analyze-data after a certain amount of episodes. Furthermore, we created a few agents who perform according to different policies in a certain market type such as:
- fixed-price agents at a linear economy environment
- rule-based agents at a circular economy environment
We also have a "humanplayer" agent so that you are able to play against competitors like in a game.
According to our project title, we introduced self-learning agents as well. They are characterized as a subtype of an artificial intelligence, that is so called Reinforcement Learning, because of gathering valuable information out of the actions that were made by customers and competitors. Hereby, RL-agents learn from this gained information and adapt their operations in order to increase their profits in contrast to the competitors.
This component of our system specifies his actions based on rule-based strategies, for instance, there are competitors who:
- calculate the quality-per-money value of each other vendor and adapt to the best determined value
- return just a random price
- set their prices in a more sophisticated manner by distinguishing between subtle quality nuances
- undercut the price of the "most valuable" vendor/agent by just a small difference and randomly raise their prices when they reached a lower profit-threshold
Online Marketplace Simulation: A Testbed for Self-Learning Agents is the 2021/2022 bachelor's project of the Enterprise Platform and Integration Concepts (@hpi-epic, epic.hpi.de) research group of the Hasso Plattner Institute.