Autonomous navigation of a mobile platform using an inverse perspective-based stereo vision system and integrated auxiliary sensors

A platform for the testing and evaluation of various sensing and motion algorithms is devel- oped in this work. Additionally, a vision algorithm suited for high-speed object detection is implemented and its integration with a ring of ultrasonic sensors is tested. Of primary interest is the development of a platform suitable for the testing of various combinations of sensor systems as well as motion algorithms. Since different motion algorithms can take advantage of different sensor modalities, the ability to select and evaluate sensor systems tuned to an algorithm is important.

To show the effectiveness of the platform for testing vision systems, associated sensors, and motion algorithms, a complete system is implemented, including a real-time vision- based object detection system. The vision system, using an inverse perspective algorithm, has been chosen for its ease of implementation and good performance. Like all vision systems, the inverse perspective vision system requires a calibration system, and a suitable one is implemented for this system. The capabilities and limitations of this vision system are discussed, as are suggestions for improving the performance in certain situations.

Finally, the combination of the vision system and a set of ultrasonic sensors is used to demonstrate the effectiveness of multi-sensor modality for obstacle negotiation. A simple motion algorithm demonstrates the use of multiple sensor modalities and the ability to plug a motion algorithm into the system. The effectiveness of the system as a test bed for sensor systems and motion algorithms is discussed, and suggestions for future work are presented.