T5 / Experimental validation
Task leader: David Hyunchul Shim (USRG, KAIST, South Korea)
The testbed vehicle is based on a Hyundai/Kia stock vehicle, Soul, which is a conventional gasoline engine-powered vehicle with electric power steering and automatic transmission. The lateral motion of vehicle is controlled by using MDPS (motor-driven power steering), which is modified to accept external steering commands via CAN network. The later controller is designed using a neural network. The longitudinal motion of vehicle is externally controlled by pushing the accelerator and the brake of vehicle using DC motor and cruise control.
This task deals with hardware (cameras, laser scanner...) installation, Code Fusion between Drive assistance system and Autonomous Navigation System (ANS) and experimental validation. The effectiveness of reconstructed image will be evaluated through scenarios for low and high speed of vehicle as follows.
Through the auto-parking scenario, the detection of parking line and obstacle can be evaluated in the situation close to the static condition of vehicle which is low speed. The detection of parking site on rear side of vehicle to parking backward and the relative position should be obtained to park into parking site even if there are any other vehicles or obstacles. Parallel Parking can be able to park in one step or multiple steps. The detection of vehicles or obstacles on lateral side of vehicle and the relative position should be obtained to park between them in enough space. In traffic-congested city, the speed of vehicle is low. In the situation, the exact distance between ego vehicle and preceding one should be obtained to avoid the collision with the preceding vehicle.
The reconstructed image should be obtained more fast than in low speed of vehicle due to its high speed. The relative position of vehicle in the straight and curved lane should be obtained. Moreover the distance from neighbour vehicles or obstacles in the front, rear and lateral of ego vehicle. The difference between the longitudinal direction of vehicle and its movement direction in curved lane can mistake the vehicle or obstacle like a fence for the preceding vehicle. It can hinder the exact collision warning.
This task is divided into two sub-tasks:
The testbed vehicle is based on a Hyundai/Kia stock vehicle, Soul, which is a conventional gasoline engine-powered vehicle with electric power steering and automatic transmission. The lateral motion of vehicle is controlled by using MDPS (motor-driven power steering), which is modified to accept external steering commands via CAN network. The later controller is designed using a neural network. The longitudinal motion of vehicle is externally controlled by pushing the accelerator and the brake of vehicle using DC motor and cruise control.
This task deals with hardware (cameras, laser scanner...) installation, Code Fusion between Drive assistance system and Autonomous Navigation System (ANS) and experimental validation. The effectiveness of reconstructed image will be evaluated through scenarios for low and high speed of vehicle as follows.
Through the auto-parking scenario, the detection of parking line and obstacle can be evaluated in the situation close to the static condition of vehicle which is low speed. The detection of parking site on rear side of vehicle to parking backward and the relative position should be obtained to park into parking site even if there are any other vehicles or obstacles. Parallel Parking can be able to park in one step or multiple steps. The detection of vehicles or obstacles on lateral side of vehicle and the relative position should be obtained to park between them in enough space. In traffic-congested city, the speed of vehicle is low. In the situation, the exact distance between ego vehicle and preceding one should be obtained to avoid the collision with the preceding vehicle.
The reconstructed image should be obtained more fast than in low speed of vehicle due to its high speed. The relative position of vehicle in the straight and curved lane should be obtained. Moreover the distance from neighbour vehicles or obstacles in the front, rear and lateral of ego vehicle. The difference between the longitudinal direction of vehicle and its movement direction in curved lane can mistake the vehicle or obstacle like a fence for the preceding vehicle. It can hinder the exact collision warning.
This task is divided into two sub-tasks:
- T 5.1 Code Fusion between DAS and ANS
- T 5.2 Experimental validation
