|
No, Title
|
SCD 6.5: AI-based near field, high resolution 360-degree perception system
|
|||
|
Leader
|
EDI
|
|||
|
Contributing Partners
|
Infineon
|
|||
|
Description
|
EDI designed, produced and assembled a modular and adjustable sensors platform for the time of flight (ToF) sensors used to create the AI-based near field, high-resolution 360-degree perception system. This sensor platform supports different ToF sensor types (support implemented for Monstar and Basler), as well as their trigger control and interfacing. A calibration algorithm and a custom fidelity check metric for inter-sensor translation and rotation were developed and introduced on this sensor platform. Implementation of vertical timestep encoding for the spiking neural networks (SNN) allows for spatio-temporal object detection by the perception system, by using spiking CNN on ToF sensor data.
|
|||
|
Deployment/utilization
|
Drivers in different vehicles, especially trucks, are susceptible to blind zones with severe safety risks. To address the challenge, EDI followed a multi-step deployment strategy. Initially, the perception system was integrated into a mobile system, which was used for early prototyping and data acquisition. Further, the system was deployed on an EDI Drive-by-Wire (DbW) vehicle to verify the developed registration and object detection algorithms. Finally, it is planned to leverage the system’s automated configuration capabilities to deploy a perception system onto Mercedes-Benz vehicles and AVL’s Drivingcube system (virtual environment) during the final project review.
|
|||
|
Pictures/visuals with titles
|
AI-based near field, high resolution 360-degree perception system’s mobile setup used for initial testing, verification and algorithm design.
|
Perception system’s integration onto EDI’s Drive-by-Wire vehicle.
Object detection based on spiking neural networks and obtained ToF data by 360-degree perception system.