Simulation of an autonomous vehicle with audvision-based navigation system in unstructuredudterrains using OctoMap

摘要

Design and implementation of autonomous vehiclesudis a very complex task. One important step on building autonomousudnavigation systems is to apply it first on simulations.udWe present here a vision-based autonomous navigation approachudin unstructured terrains for a car-like vehicle. We modelled theudvehicle and the scenario in a realistic physics simulation with theudsame constraints of a real car and uneven terrain with vegetation.udWe use stereo vision to build a navigation cost map grid basedudon a probabilistic occupancy space represented by an OctoMap.udThe localization is based on GPS and compass integrated withudwheel odometry. A global planning is performed and continuouslyudupdated with the information added to the cost map whileudthe vehicle moves. In our simulations we could autonomouslyudnavigate the vehicle through obstructed spaces avoiding collisionsudand generating feasible trajectories. This system will be validatedudin the near future using our autonomous vehicle testing platformud- CaRINA.