Korean electric vehicle producer IT Engineering, and The Electronics and Telecommunications Research Institute (ETRI), have partnered on the development of a smartphone package that enables users to call and move a self-driving vehicle through voice recognition, according to ETRI. Commands as simple as, “start” and “go” may one day drive autonomous vehicles into the future.
Utilizing base levels of 0 to 5, as suggested by the U.S. automotive technology association, developers tested this unique technology at a Level 3 (eyes off the road) and Level 4 (mind off the road). Through extensive research and testing, the innovative software showcased how it allowed a self-driving vehicle to automatically produce and update an accurate map of its surrounding environment. With an error range of less than 10 cm, this groundbreaking technology proves that we are well on our way to fully functioning autonomous vehicles.
The application requires sensors, which recognize the surrounding environment on the road, and deliver precise mappings of the area—including traffic lights, boundaries, landmarks, and various other components that come with being on the road. In order for the application to be a success, it heavily relies on precise mapping for real-time results and monitoring.
In addition to the creation of such technology, researchers were able to identify the electrical impact of the vehicles once put to the test. Current vehicles being tested for autonomous driving are midsize to large sedans, which require hundreds of watts of electricity, due in part to the significant amount used in the sensor and artificial intelligence (AI) algorithm operating processes.
Staggeringly, ETRI has been able to lower the electric consumption to below 100 W, which is the same usage as two notebook laptops. According to ETRI, their success is due to its use of compact vehicles instead of large SUVs. Testing continues for ETRI, as they forge paths to discover new ways to improve the software algorithms and achieve even better results through deep learning.