On a robotic Matlab Java testbed for mobile robots

Sammanfattning: This thesis presents the MICA (Mobile Internet Connected Assistant) testbed and how it can be used to interface a mobile robotic system; in this case a modern wheelchair equipped with a CAN bus. In MICA we do research in embedded Internet systems, sensor technologies, navigation algorithms and in the knowledge about user. In order to incorporate this we have developed the MICA testbed. The MICA wheelchair has sensors like laser scanners, cameras, rate gyros, odometric encoders and flash cameras. The MICA testbed software can be divided into three main parts; the embedded server software, the Java clients and the MATLAB algorithms. The client software gives the user access to the sensor databases. The clients are written in JAVA which gives the code portability and the possibility for use in MATLAB. MATLAB is a good environment to do research of new algorithm in; as the algorithms work, they are step-by-step ported to Java. The MICA wheelchair has an embedded PC that runs Linux, which is used to collect sensor measurements continuously and store them into databases. All sensor databases can easily be accessed through the MICA software. It is possible to access and manipulate the wheelchair and its sensors over Internet since all sensor measurements are time stamped and stored in the system. The MICA client/server solution with Wave-LAN (IEEE802.11) provides the possibility to write complex distributed programs that runs on a network. Partly analyzed and processed data, for instance by a MATLAB script, can be sent to other clients on the network for further use. Lots of algorithms are written during research and the software becomes encapsulated and hard to maintain for other people than the programmer if they are written in programming languages like C. In MICA, new algorithms are mainly coded in MATLAB which makes it easy for other people to reuse and contribute in the developing process of new algorithms. Tests have shown that the MICA testbed is easily understood and used by undergraduate students in education. The MICA testbed is also a good environment to do research with. As the sensor are directly available to arbitrary clients over the Internet the system is ready for multi-vehicle implementations. %One vehicle can access sensor measurements from other vehicles; for instance odometric, GPS data or range laser scans. This makes it possible to evaluate algorithms that uses several platforms. The thesis also presents a navigation filter used to estimate 10 states of an airborne UAV; these were position, velocity and attitude. The Kalman filter was written in extended information form and used a quaternion to represent attitude. The observations feed to the filter were position, velocity and attitude. The estimated errors were feed back to the INS. A setup with four GPS receivers were used to calculate the attitude of the airframe. The coming research in MICA will focus on navigation algorithms using cameras and laser scanners.