Self-learning roadside radar to prevent wildlife accidents
Professor Dr. Hubert Mantz from Ulm University of Applied Sciences, one of the heads of the “Salus” project, explains the background: “Car manufacturers are installing driver assistance systems, starting with high-end models, but it will take a long time before such aids are available in all vehicle models. Similarly, assistance systems are now being installed in high-end motorcycles, but lack of space makes it difficult to reach the level of car-based systems there. The aim of our project is to install small sensor units at the roadside that can detect hazards and report them to approaching vehicles. In addition, for road users without reception systems, street lighting can be switched on to highlight the danger area and/or warning signs can be activated. Our system detects and warns of the hard-to-detect dangers on country roads and should significantly improve traffic safety”.
The current demonstration system can process data from three sources simultaneously – radar, optical cameras and infrared cameras. Additional sensors could be integrated to measure air pollution levels, for example, which has already attracted commercial interest from companies in the project consortium. The Salus project envisages the use of the stand-alone units on roads in Germany, mounted on piles at the roadside, which means that they must be cost-effective and solar-powered.
The communication system between the units, which will be installed at the roadside, must also be designed for low power so that it can be powered by solar cells. A LoRaWAN (Long Range Wide Area Network) is therefore used, as this standard is characterised by low power consumption and is based on licence-free frequency bands, so that no further costs are incurred. A LoRaWAN provides a range of up to 40 km in rural areas, which is more than sufficient for the installed units to function together as one large warning system.
The system itself learns by means of neural networks so that it can distinguish between e.g. cyclists, cars or deer. According to Mantz, the capabilities of the system should go far beyond the mere detection of movement.
Currently, the project is in a critical phase, namely the classification of recognized objects. According to Mantz, this has never been done like this before. Through classification, the system should be able to predict the movements of the objects, which provides extremely useful real-time information to make a very accurate prediction of how a dangerous situation will develop.
During development, the researchers are using Spectrum’s PCIe digitizer M2p.5926-x4, which offers 16 bits, 10 MHz bandwidth and four differential inputs. It collects all the information our self-learning system needs and can process signals from all sources simultaneously in real time. In addition, Mantz says, the operation of this digitizer is very simple and intuitive, allowing researchers to focus on the project rather than spending time programming.
Related articles:
LoRaWAN wireless accelerometer with edge computing
LoRaWAN ‘sees’ growth of over 60 percent in 2018 covering nearly 100 countries
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
