Armory Siege

The team defends a central core against waves of enemies while rotating through gated weapon, ammo and medical rooms. Agents must split between holding the core and resupplying.

import comrad

env = comrad.parallel_env("armory_siege")

Gymnasium id	comrad/armory_siege-v0
Agents	2 by default (supports 2–8)
Capability tested	Combat coordination, resource routing, rotating roles
Episode limit	4500 steps
Observation	RGB image (72, 128, 3) per agent; measurement vector available with additional_input=True
Action space	Tuple(Discrete(3), Discrete(3), Discrete(3), Discrete(2), Discrete(3))
Default shared_reward	0.0: each agent receives its own reward
Score (info["true_objective"])	Shaped episodic return

Actions

Each agent sends one choice per head; index 0 is always "do nothing".

1. move: none / forward / backward
2. strafe: none / right / left
3. turn: none / left / right
4. attack: none / attack
5. weapon: none / weapon 1 / weapon 2

Reward

Per agent: bonuses for kills, hits, new weapons, ammo, healing and surviving the timeout; penalties for core damage, core loss, death and firing with no ammo. true_objective is the episodic shaped return.

The score in info["true_objective"] is what the benchmark reports. The shaped reward is only a training signal.

Variants

• comrad/scenarios/batch_armory_siege/ and batch_armory/: layout variants for generalisation tests.