Model-based Testing of Real-time Adaptive Motion Planning (RAMP)

It is a practically challenging problem to test the functionality of autonomous systems and assess their performance in environments with unknowns and unpredictability. Existing testing techniques are designed heavily based on testers experience and hardly take into account all scenarios. This paper applies a model-based testing technique to evaluate the functionality and performance of a Real-time Adaptive Motion Planning (RAMP) system. First, RAMP components and their interactions are tested. Next, the whole RAMP system is tested against mobile obstacles of unpredictable motion. The model-based testing technique models the behaviors of RAMP components and the mobile obstacles with a Communicating Extended Finite State Machine (CEFSM). The behavior models are then leveraged to generate Abstract Behavioral Test Cases (ABTCs). These ABTCs are subsequently transformed by test data into executable test cases. The generated executable test cases are further applied to a RAMP implementation based on the Robot Operating System (ROS) and executed with the software testing tool, Google Test. The results demonstrate an effective use- case of applying a systematic software testing technique to the evaluation of real-time robotic systems.