|
The development of artificial intelligence is taking place at a high pace on devices with large computing resources, which are constantly increasing data storage volumes and improving performance. However, attention to devices on the periphery seems to be low due to the lack of computing resources. But such devices also have an important advantage - a large physical quantity, which can be used in building distributed systems with artificial intelligence. In this regard, this article is aimed at identifying the possibilities of using peripheral devices in complex, including distributed information systems. Three approaches to such systems are analyzed and described: with and without peripherals, and an integrated approach when peripherals form an auxiliary role for a complex system. This approach involves the introduction of machine learning models into the control device and the collection device. For the approach, the situation of application in an unmanned vehicle will be simulated, where road signs are the object of detection. An experiment was conducted for the simulated situation, the results of which confirm the expediency of using an integrated approach for a system with artificial intelligence. Based on the results, the main advantages and disadvantages of implementing an integrated approach using peripheral devices in complex systems are highlighted. This work will be useful for researchers in the field of the Internet of Things, artificial intelligence, as well as specialists in the development of unmanned vehicles.
Keywords:Edge Computing, Edge AI, TinyML, distributed systems, Tiny AI, self-driving cars, MEC
|