Software Stack for Autonomous Vehicle: Motion planning

An autonomous vehicle can operate itself without any human interaction through the capacity to sense its surroundings. It requires a navigation plan to travel avoiding collision and following all the traffic rules at the same time. Navigation issue is a computational issue to discover the succession of legitimate designs that move the object from source to destination. In the robotics field, this term is known as motion planning or path planning. There exist well-recognized methodologies for this problem; however, by applying some helpful heuristics, a better version of driving API can be designed. This study develops a complete software architecture needed for autonomous vehicles. It gives brief insights about available techniques for each module involved in motion planning and possible optimizations to achieve better results. The global planner uses a road network stored in open drive format to find themost eligible global path for ego vehicle. Whereas reactive local planner uses surrounding information to plan better paths avoiding static and dynamic obstacles. At last, the results are introduced in examination with existing methods to show improved measurements accomplished by this framework.