SpiceMerchant

A GOAP (Goal Oriented Action Planning)(based on Finite State Machine) AI system case implementation from the modern game algorithm course.

General:

A brief description of what this project is actually demoing

In this system, there is one spice merchant and eight traders doing business following certain rules. The ultimate goal of the player-agent is to store 2 of each spice in the caravan. Spices are acquired by trading different 20 spices between the traders. Each trader accepts a type of spice and returns another type.

• Trader 1: Gives you 2 turmeric units.
• Trader 2: Takes 2 Turmeric units and gives you 1 Saffron unit.
• Trader 3: Takes 2 Saffron units and gives you 1 Cardamom unit.
• Trader 4: Takes 4 Turmeric units and gives you 1 Cinnamon.
• Trader 5: Takes 1 Cardamom and 1 Turmeric and gives you 1 Cloves unit.
• Trader 6: Takes 2 Turmeric, 1 Saffron and 1 Cinnamon and gives 1 Pepper unit.
• Trader 7: Takes 4 Cardamom units and gives you 1 Sumac unit.
• Trader 8: Takes 1 Saffron, 1 Cinnamon and 1 Cloves unit and gives you 1 Sumac unit.

The player-agent begins next to the caravan, and has an inventory that allows carrying at most 4 units of spice at any one time. The caravan has unlimited capacity, and traders have unlimited supply.

A quick demonstration:

To set the goal as "having 2 of Tu(Turmeric) and Sa(Saffron) at the end", we do the following implementation: (Similar as the worldData implementation, which we have to listen and update each frame)

Then, through our GOAP system, it should return a full plan telling us what concrete actions we should execute in sequence to satisfy the goal action.

To take a deep look at the first action we are gonna take from the correcponding file, basically we have to move to the trader 1's position and update the number of Tumeric from the backpack component.

The standing positions for all traders are generated randomly at runtime. We can see that the red cylinder (stands for the merchant here) is moving to the Trader 1 first, the Text UI updated itself as well.

Then Trader 2 and we successfully got 1 Sa by giving out 2 Tu.

Then the merchant goes back to the caravan and drop 2 Tu to satisfy the first goal, then it will be taking the rest of the actions to satisfy the second action.

The NavMesh navigation system from Unity was used to move agents.

Source Code:

• Part of the GOAP code skeleton is implemented by @sploreg, I modified some detailed for this specific case.