Development of autonomous vehicle navigation in unstructured environments: the impact of implementing a path planning algorithm on autonomous vehicles

The path planning system has been identified as an efficient method for optimizing navigation, reducing energy consumption, and ensuring safety in autonomous vehicles. Various studies have been conducted on algorithms such as Ant Colony Optimization (ACO), Artificial Potential Field (APF), and the A* algorithm. However, only a few studies have evaluated the effectiveness of each of these algorithms, especially for autonomous vehicles implemented in real-road scenarios. Thus, this study aims to assess the effectiveness of ACO, APF, and A* in identifying the optimal path, computational efficiency, and execution time for generating routes in an autonomous vehicle. Experiments were conducted using road coordinate data from the Universitas Sriwijaya campus, representing suburban road conditions in Indonesia. The results showed that the A* algorithm excels in finding optimal routes, with an average path length of 0.48 km and a 100 % success rate. This is due to its heuristic, Euclidean-based approach. Meanwhile, ACO achieved an average path length of 0.57 km with a 100 % success rate, whereas APF achieved 0.36 km with a 41 % success rate. ACO demonstrated varied route performance due to its probabilistic nature, while APF generated paths more quickly but often failed in complex environments due to local minimum traps. Regarding computation time, an increase in distance leads to a longer route formation time for APF and A*, respectively. However, in ACO, route distance does not directly determine the time required for route formation, as the algorithm incorporates a probability factor in the process. This study confirms that A* is more optimal for global path planning, whereas APF is better suited for local path planning. These findings provide valuable insights into the development of autonomous vehicle navigation in unstructured environments