Alexandria Engineering Journal (Aug 2023)
A transfer-learning-based energy-conservation model for adaptive guided routes in autonomous vehicles
Abstract
Autonomous vehicles (AV) are expected to improve road safety and reduce traffic congestion by optimizing routes and reducing human errors. AVs have the potential to increase accessibility for people with disabilities and reduce the environmental impact of transportation. AVs require radio transmitters to communicate with other vehicles and infrastructure, external charging to power their electric motors, and communication equipment to receive real-time data about traffic and road conditions. Additionally, these requirements must be met for AVs to operate efficiently and conserve energy. Therefore, this work introduces a novel technique called energy-conservation guided route adaptation (EC-GRA) that aims to enhance the energy efficiency of connected vehicles. With the balance in energy adaptation for distinct purposes, the utilization rate is adjusted for communication and navigation. The complex decisions are confined to the energy availability and conservation factors required in an adaptive driving condition. This technique employs transfer learning to update the available and adaptable energy ratios under displacement-based route adaptations. In the learning process, the transfer and update states for displacement-aware decisions under varying scenarios are modeled. This study validates the state transitions involved in recommending energy utilization during both autonomous and guided driving scenarios. The results show that the proposed methodology exhibits superior performance compared to the currently available techniques. The EC-GRA under consideration has demonstrated an average energy conservation ratio of 45.58. The decision rate for this method is 0.63/navigation, while its energy utilization is 126.37 Joules. The number of failures observed in the proposed EC-GRA is 6/navigation, which represents an improvement over the existing approach.
