One benefit of formal verification is the removal of design errors at early stages of complex artifacts. This is being used extensively for software with great success. We wish to extend this to domains where the distance between formal models and the real application is bigger. We choose to work on robotics. For this field, there is a large distance between models and real implementations, because our knowledge of physics is incomplete, the world needs to be represented through geometrical abstractions, and there are limitations to what computers can compute accurately and what sensors and actuators can do in terms of precision. To begin with we will look only at a question involving geometry. We wish to present a comprehensive Coq program to compute trajectories for a point between obstacles that are given by straight line segments.