The system consists of control software and embedded equipment for unmanned electric vehicles based on a combination of evolutionary (immune) algorithms and artificial neural networks with the ability to learn to avoid collisions without human intervention. In the finished version, the product is an embedded computer (microcomputer) with a connected positioning system module (e.g. GNSS), a wireless communication module (4G, 5G, Wi-Fi or other radio frequency) and a camera system. Different operating modes are possible: informative (can work in vehicles with a human as an auxiliary system), active (requires connection to the actuator controls), autonomous (for unmanned vehicles). The system equipment in individual vehicles works to communicate with each other to predict the probability of a collision and make the best decision to minimize it. Unequipped obstacles are detected by the camera and the computer vision unit. When working in a vehicle, the system itself will learn faster and better to find solutions while always improving operational efficiency. The technology and software have been developed, but the system needs to be prototyped and tested in real conditions.
The system can be used in any vehicle (rail, road, water, air) and mobile robot systems (indoors, warehouses, etc.) to increase traffic/moving safety. Without connection to the vehicle's actuators, the system is completely autonomous and can act as a driver assistance system. The best efficiency can be achieved in autonomous mode and in environment where several vehicles are equipped with such devices and form a data exchange network, rather than just using a computer screen to identify objects / obstacles. Thus, the system is intended to ensure the collision-free operation of a team of mobile robots or vehicles.
Autonomous system. As a driver assistance system, only connection to the power supply is required. In any mode does not require human input (although optionally a person could adjust the sensitivity), adapts itself to any vehicle, to particularities of acceleration / braking / trajectory changing, constantly evolves itself by learning to make better solutions to minimize the collision probability as quickly as possible in real time.
artificial intelligence,drones,evolutionary algorithms,machine learning,neural networks,vehicle,transport,unmanned vehicle,collision prevention,anticollision system,traffic safety,computer vision