Local and global path generation for autonomous vehicles using splines

摘要

Autonomous vehicles soon will be a reality for many daily situations. Currently, the main barrier preventing the development of autonomous vehicle is its high cost. In agreement with the objectives of the Autonomous Mobility Laboratory from the Mechanical Engineering Faculty at Unicamp of developing inexpensive autonomous robots, the presented article exposes relevant aspects from the implementation process of a real time path planning technique for terrestrial vehicles. The studied technique can be divided into four stages, which are: global route construction, car localization, paths generation and path selection. The construction stage plays an important role in the robustness of the presented technique. The adequate number and distribution of waypoints are highly correlated to well global route construction and to a satisfactory arc length reparameterization. Unfortunately, the actual aspects of the waypoints provided by on-line services are inadequate to its directly usage in the construction of the global route curve, making necessary its prior processing. There is also a trade-off between the numbers of waypoints and the quality of the desired arc length parameterization that needs to be tuned.