Advanced IoT-based Algorithms for Improved Object Detection and Navigation in Autonomous Vehicles

In recent years, the progression of communication, information and sensor technology applications has proliferated the autonomous vehicle market. Ensuring the safe and comfortable driving of autonomous vehicles is crucial for maintaining safe interactions between Autonomous Vehicles and pedestrians, and for mitigating accidents. However, autonomous vehicles face several technological challenges, including poor accuracy in diverse object detection on the road and poor driving comfort. This paper proposes an Internet of Things-enabled novel Tabu Search-based Chaotic Dragonfly Optimization algorithm combined with a Multi-Hypergraph-based Compressed Gated Recurrent Unit and a Bidirectional Recurrent Neural Network-based Double Deep Q-Learning framework, aiming to ensure the safety and comfort of Autonomous Vehicles in real-time driving scenarios. In the initial phase, the Tabu Search-based Chaotic Dragonfly Optimization algorithm is constructed for feature selection, selecting optimal attributes, and providing effective detection performance. In the second phase, a Multi-Hypergraph-based Compressed Gated Recurrent Unit framework is implemented for detecting abnormalities on the road. This framework incorporates an Asymmetric Convolutional Layer to reduce the model's parameters and computational requirements while maintaining spatial information. In the final phase, a Bidirectional Recurrent Neural Network-based Double Deep Q-Learning Network is introduced for learning navigation decisions in dynamic environments and controlling Autonomous Vehicle speed. This framework includes an attention mechanism, which captures long-term dependencies and enhances learning efficiency. The proposed model achieved a higher accuracy of 98.95% and a low average delivery latency of 88 milliseconds at a vehicle speed of 70 km/h. Overall, the proposed model effectively reduces latency, enhances detection accuracy, and improves driving comfort and safety.