English | 中文
RideSafeUM is an initiative co-funded by EIT Urban Mobility that brings micromobility safety benefits to users, public authorities and operators through the use of innovative technology. This initiative proactively prevents riding incompliances, provides smooth reaction in case of accidents, and gathers data for wider learning, policy-making and management purposes. The project was born in response to the need of making micromobility safer, and to encourage increased ridership of what have become key modes for the future of our mobility systems.
The RideSafeUM solution is based on the integration of computer-vision software, with camera, GPS and an accelerometer. This technology has resulted in an app which can be either supported by users’ smartphones or through external equipment. Versatility being one of the core values of RideSafeUM, the app is available to both private riders and to be easily integrated with shared micromobolity operators’ apps. On the background, a city dashboard enables authorities to set the restrictions and warnings displayed in the users’ app, thereafter being able to identify and dynamically manage micromobility safety issues.
“The system works on a bi-lateral communication basis. Real-time digital information of regulations is displayed to the user (via the app or the operator’s front-end). At the same time, anonymised alerts are sent to the authorities if an accident occurs, using a black-box function to identify safety patterns.”
This is an accelerometer-based accident detection for electric scooters (E-Scooters) developed on the basis of anomaly event detection. I uploaded several simple detection algorithms for users to study. The basic algorithm is very important, which is of great help and reference for subsequent optimization and development. In AccidentDetection_CFAR.m, the acceleration signal is analyzed by the CFAR algorithm to detect and identify accident events. Based on this, other detection algorithms is developed that process and analyze the data and signals to detect anomaly events (crash or accident).
For the basic principles and framework of the CFAR algorithm, you can refer to my repository of CFAR-based SAR ship detection, or refer to the following MATLAB official web site: Constant False Alarm Rate (CFAR) Detection
Among them, AccidentDetection_Realtime.m algorithm simulates real-time accident detection to realize traversal of all data. It will stop traversal when an abnormal situation is detected and judges whether it is an accident, then continues to traverse the remaining data after the detection. In this case, the funtion of circshift in MATLAB is used to simulate the process of acceleration data being monitored in real-time.
Three test data are uploaded to the data folder for studying. Two of the files where the filename contains "crash" have two accident events and additional "disturbances" (steps, violent shaking, etc.). The Normaldrive file is a part of normal drive data with U-turn behavior for comparison.
Algorithm CFAR:
Algorithm 2:
Algorithm 3:
Implementation of Signal Processing Algorithm on a smartphone: ISPA