Multi-Agent-Reach-Avoid-Game

Hamilton-Jacobi (HJ) reachability analysis based multi-agent reach-avoid game repo.

This repository contains the implementation of the paper "Multi-Agent Reach-Avoid Games: Two Attackers Versus One Defender and Mixed Integer Programming (2023 CDC)". This implementation is based on the OptimizedDP library developed by the MARS lab, Simon Fraser University.

Step 1: Set Dynamics

The dynamics of attackers and defenders (including the maximum speed) in different reach-avoid games are in files AttackerDefender1v0.py, AttackerDefender1v1.py and AttackerDefender2v1.py.

Step 2: Generate Value Functions

We need to calculate several value functions for the further reach-avoid games.

• Run the file MRAG/hjvalue1v0.py to obtain the 1 vs. 0 value function 1v0AttackDefend.npy, which is used to generate control inputs of attackers.
• Run the file MRAG/hjvalue1v1.py to generate the value fucnton 1v1AttackDefend_speed15.npy, which is used to generate control inputs of the defender in the 1 vs. 1 reach-avoid game.
• Run the file MRAG/hjvalue1v1_basement.py to generate the value fucntion 1v1AttackDefend_g30_speed15.npy for the computation of the 2 vs. 1 value function.
• Run the file MRAG/hjvalue2v1.py to generate the value function 2v1AttackDefend_speed15.npy, which is used to generate control inputs of the defender. This procedure will take about 1.5 hours and consume around 20 gigabytes of RAM (tested on a 16-thread Intel(R) Core(TM) i9-9900K CPU at 3.60GHz).

So far, all necessarcy value functions are calculated.

Step 3: Run the Games

3.1 Reach-Avoid Set Visualization

Run the file MRAG/plot_figures.py to obtain the figures of reach-avoid set 2v1_RAS.png.

3.2 The 2 vs. 1 Reach-Avoid Game Simulation

• Run the file MRAG/game2v1.py to obtain the results of our proposed method. To get the positions of different time instances, change the corresponding setting of the total game duration T:

T = 0.785 # attackers_stop_times = [0.475s (95 A1 is captured), 0.69s (138 A0 by D0)]

3.3 The 6 vs. 2 Reach-Avoid Game Simulation

• Run the file MRAG/game6v2.py to obtain the results of our proposed method. To get the positions of different time instances, change the corresponding setting of the total game duration T.

• Run the file MRAG/game6v2_basement.py to obtain the results of the baseline method. To get the positions of different time instances, change the corresponding setting of the total game duration T.

3.4 The 8 vs. 4 Reach-Avoid Game Simulation

• Run the file MRAG/game8v4.py to obtain the results of our proposed method. To get the positions of different time instances, change the corresponding setting of the total game duration T.

• Run the file MRAG/game8v4_basement.py to obtain the results of the baseline method. To get the positions of different time instances, change the corresponding setting of the total game duration T.