Collision Avoidance Of Arbitrary Polygonal Obstacles

摘要

Systems, methods, and computer readable media to improve the operation of graphics systems are described. In general, collision avoidance techniques are disclosed that operate even when the agent lacks a priori knowledge of its environment and is, further, agnostic as to whether the environment is two-dimensional (2D) or three-dimensional (3D), whether the obstacles are convex or concave, or whether the obstacles are moving or stationary. More particularly, techniques disclosed herein use simple geometry to identify which edges of which obstacles an agent is most likely to collide. With this known, the direction of an avoidance force is also known. The magnitude of the force may be fixed, based on the agent's maximum acceleration, and modulated by weighting agents

主权项

1. A collision avoidance method, comprising:
identifying a current position of an agent in an environment; determining a motion vector corresponding to the agent's current position; identifying a first obstacle in the environment, the first obstacle having a plurality of edges; determining a first plurality of edge vectors, each edge vector corresponding to an edge of the first obstacle and extending between a closest point of a corresponding edge and the motion vector, wherein each of the first plurality of edge vectors is orthogonal to its corresponding edge at the corresponding edge's closest point; selecting, from the first plurality of edge vectors, a first edge vector that intersects the motion vector closest to the agent's current position and whose corresponding closest point is within a bounding radius of the agent; determining a first force based on the first edge vector and the motion vector; and applying the first force to the agent.