Tactical Reinforcement Intelligence for Swarm Hunting and Unmanned Logistics
Project TRISHUL is an autonomous, defense-grade Space Situational Awareness and UAV swarm coordination architecture. It is designed to operate in highly contested, GPS-denied, and bandwidth-constrained environments.
POMDP Bayesian Filtering: Agents maintain probability distributions of hidden enemy targets rather than relying on perfect radar coordinates.
Byzantine-Resilient GNN: The swarm uses a Graph Neural Network with Median-Aggregation to mathematically ignore false telemetry broadcasted by hacked (rogue) drones.
Neural Lyapunov Control Barrier Functions (NLCBF): A mathematically verified safety shield that guarantees zero collisions and uses Lyapunov attraction vectors to prevent symmetric deadlocks during evasive maneuvers.
Information Bottleneck (Radio Silence): Drones calculate the kinematic entropy of their flight paths and only transmit data during unpredictable maneuvers, drastically reducing bandwidth and SIGINT detection.
Explainable AI (XAI) via SINDy: Extracts transparent, human-readable governing physics equations from the neural network's latent space using Sequential Threshold Ridge Regression.
1. Clone the repository: bash git clone https://github.com/your-Aemili09/TRISHUL_Independent.git
2. Install the required dependencies:
pip install -r requirements.txt3. Run the tactical dashboard:
python project_trishul.py