Approaches to speed planning for ground-based autonomous vehicles

The article provides an overview of speed profile planning methods for ground-based unmanned vehicles. Autonomous driving technology is in the active development phase, and many tasks have already been solved, but with insufficient quality for safe operation and scaling of the technology. Thus, this field is promising from a scientific and engineering point of view. The paper examines in detail the formulation of the problem, describes various types of constraints to the final solution, and considers the construction of the task's functionality, depending on the requirements and methods used. Special attention is paid to the ST graph as a tool for modeling interaction with other traffic participants. Practical scenarios of agent interaction in various situations that often arise in traffic are also analyzed in detail. The authors focus on a class of methods for constructing a velocity profile along a known geometric trajectory. Two families of approaches are considered: dynamic programming and reducing a continuous problem to a quadratic programming problem. The existing modern methods based on these approaches are analyzed in detail, and improvements and improvements are proposed in the context of various additional requirements for the solution. The article serves as a methodological basis for developers of autonomous driving systems.