University of Michigan researchers are teaching self-driving cars to recognize and predict pedestrian movements with greater precision by studying in on humans’ gait, body symmetry and foot placement.
Researchers have created a “biomechanically inspired recurrent neural network” that can predict poses and future locations for one or several pedestrians up to about 50 yards from the vehicle.
The network catalogs human movements using data collected by vehicles through cameras, LiDAR and GPS to capture video snippets of humans in motion and then recreates them in 3D computer simulation. To create datasets researchers parked a vehicle with Level 4 autonomous features at several Ann Arbor intersections.
The network can study the pace and periodicity of human movement:, the mirror symmetry of limbs, foot placement affects to equip vehicles with the necessary predictive power.