In static ad-hoc routing, dynamic topological changes like node mobility easily break paths, requiring expensive route re-discovery. Under opportunistic routing, multiple nodes that hear the same broadcast transmission may attempt to forward the packet simultaneously, causing duplicate packets and network congestion. Efficiently coordinating which single node forwarder should proceed is the primary challenge.

We study dynamic coordination mechanisms where potential forwarders self-organize to select the optimal relay based on geographic proximity to the destination, signal strength, and remaining battery levels. For extremely sparse networks where connections are frequently interrupted, we integrate Delay Tolerant Network (DTN) routing to store data locally and physically carry it via mobile nodes to guarantee message delivery.

Human visual and cognitive comfort follows the law of diminishing marginal utility relative to physical resource allocation, meaning that boosting bandwidth beyond a certain point does not improve user satisfaction. In addition, QoE is highly contextual, varying based on video scene dynamics, gameplay tension, and individual user expectations, making static resource allocation suboptimal.

We capture biological feedback such as EEG and eye-movement signals to dynamically assess visual attention and cognitive stress, allowing the network to prioritize bandwidth for critical visual frames. For cloud-hosted multiplayer games, our system analyzes game-state metadata to forecast active combat zones and dynamically schedules routing priorities to eliminate lag for engaged players.

Drones suffer from strict battery capacity limitations and restricted sensor footprints. The fleet must coordinate to solve a complex trade-off: maximizing coverage to locate survivors within the critical golden hour post-disaster, while maintaining mutual wireless connectivity and avoiding spatial collisions under highly dynamic topologies.

We implement bio-inspired swarm heuristics modeled after locust behavior and game-theoretic bidding algorithms, allowing individual drones to dynamically balance search and relay roles based solely on local measurements and neighbor interactions. Our work also extends to trajectory optimization for UAVs acting as physical data carriers (Data Mules) to retrieve information from ground sensors, along with wireless power transfer scheduling to maximize network lifespans.

IoT devices possess severely limited computational and energy resources, preventing the deployment of heavy encryption or active endpoint detection software. Consequently, traditional signature-based intrusion detection systems and static machine learning models that monitor nodes in isolation fail to detect unknown zero-day exploits or stealthy botnets that carry out coordinated low-rate attacks across multiple terminals.

We develop non-intrusive intrusion detection using Graph Neural Networks (GNNs) to trace relationship graphs and detect suspicious patterns across networks. Our work also implements active probing using IPID headers to estimate remote traffic safely, secure log recording via processor-level Trusted Execution Environments (TEEs), and collaborative federated learning models to detect zero-day anomalies.