MIT and Motional translate the decisions of a driving system into comprehensible concepts, tested on private tracks and in simulation.
On September 2, MIT presented CW-Net, developed with Motional to explain the decisions of an autonomous vehicle. The system combines the operation of the planner with accessible notions, such as proximity to a cyclist, without changing its driving performance.
The explanations helped safety drivers to better predict the vehicle's behaviour on a private track. A simulation study with non-specialists obtained results in the same direction, and the method could also make it easier for engineers to diagnose errors. However, these experiments do not show a reduction in accidents in real traffic.
Read the full article on the original source
Go to the source →