Map Building Intelligent Sensor for an Autonomous Vehicle

Abstract

In this paper the results in the development of a sonar array sensor for the use in mobile robot platforms are presented. The work involves hardware development and signal processing techniques for the reconstruction of the scene. The developed sensor uses an array of four transducers to improve the distance measurement and angular resolution. Its angular response has some sidelobes around the main lobe. The main advantage of this approach is a sharper main lobe, compared with the main lobe of the only one pair emitter-receiver device. However, the lateral sidelobes have significant amplitude and can produce false peaks detection. This problem can be addressed using adequate deconvolution methods. This approach decreases the amplitude of the sidelobes and produces a high quality data with low angular ambiguity. This allows map building with better data, obtaining good results in real environmentsThe tests performed show good enough obstacle detection results in a normal office environment. The signal processing treatment is adequate for the DSP implementation allowing the use of the sensor on a real-time mobile vehicle. The use of the measurements in sensor fusion levels for the map building, reduces the complexity and uncertainty of this process, allowing a simple signal-fusion level and increasing the quality of the final map