Embedded perception concept integrated into Easymile driverless shuttle

In research conducted under the PULSE program’s Perfect project, IRT Nanoelec is developing high-added-value embedded perception technologies for smart mobility, especially in dynamic environments. The capabilities developed include fusing data from multiple sensors, filtering and interpreting data in time and space, detecting and tracking moving obstacles, and identifying and locating potential collision hazards.
The technologies were originally developed for the automotive industry and integrated onto mobile architectures. Here, they have been adapted to respond to other smart-mobility-related challenges.
Easymile and IRT Nanoelec joined forces several months ago on a proof-of-concept prototype for the integration of environmental perception technologies. The prototype was embedded on an NVIDIA Tegra CPU directly connected to the shuttle’s on-board system. The initial results were conclusive and the partners plan to continue to work together.

A screenshot from the system: a cloud of data points is classified with the ground in green, obstacles (a pedestrian behind the shuttle and another vehicle straight ahead) in purple.