Optimizing AI-Driven Task Management in Autonomous Driving: A Multi-Objective Offloading Approach for Automotive Consumer Electronics

The rapid advancement of AI and the consumerization of technology have significantly reshaped the automotive industry. The integration of large models (LMs) can enhance the functionalities of vehicles such as autonomous driving, real-time decision-making, and driver status monitoring. However, this integration presents significant computational challenges that often exceed the processing power of in-vehicle systems. This study proposes a novel method for partitioning complicated computing tasks into smaller, interdependent subtasks, which are then offloaded to Mobile Edge Computing (MEC) servers to optimize the overall performance of the system. The proposed Hybrid Adaptive Multi-Objective Genetic Algorithm (HAMOGA) is introduced to manage this offloading process, optimizing both latency and energy consumption. HAMOGA dynamically adjusts the crossover and mutation probabilities based on individual fitness and iteration count. Four distinct structures of DAGs with varying scales are established for evaluating the performance of our method. Experimental results demonstrate that HAMOGA consistently outperforms established methods in various task scales. The algorithm achieves excellent performance in optimizing latency and energy consumption while maintaining robust convergence characteristics. Our method provides a practical solution for AI task offloading in intelligent vehicles, enhancing the value of automotive consumer electronics.