Modular Design & Control of a Rreconfigurable Multi-Agent Robotic System for Urban Inspection
By JEAN-PHILIPPE CLERC
December 2003
Chair: Dr. Gloria J. Wiens
Major Department: Mechanical and Aerospace Engineering
Reconnaissance, survey and inspection are critical areas that can lead to catastrophic consequences if they are not accomplished effectively. In order to minimize human error and reduce cost, an autonomous inspection multi-agent robotic system is needed. The robot should be able to perform effective inspection as well as posses climbing abilities. Most modes of mobility are either efficient on flat surfaces or efficient climbers. There remains the need to develop a dual purpose robotic system that is an effective climber and effective in coverage of flat horizontal areas. The approach for the design and analysis of a robotic system that will meet the above objectives is considered through two innovative designs. Reconfigurable and modular approaches combined with a smart structure are considered to achieve this goal and provide access to a wide range of areas.
The research focuses on eliminating mobility constraints inherent in current designs, thus extending the robot’s adaptability and range by providing the robotic system mixed/multiple mobility modes. It is assumed that the robots will navigate into a known and controlled environment equipped in order to allow the free movement of the robot (e.g., smart structure). The mechanical designs, kinematic analysis, sensor suite of the system are presented as well as testing and results.
Two Hilare type robots were designed and custom built in order to test the path planning, inspection and docking algorithms. Random with one and two robots, teleoperated and semi-approximate cellular decomposition technique were tested and compared in terms of percent coverage of a predetermined area with and without obstacles.
In order to access elevated area the robot uses a modular approach to climbing, two similar robots dock together, back to back, in order to become a serial-paired climbing robot. The system requires only two modules to perform climbing where each individual module is an autonomous robot able to perform inspection. The recursive Newton-Euler algorithm was used to calculate the torque at the four joints of the climbing robot. Due to the weight limitation and actuation restriction, an algorithm for computing path for climbing using the lowest possible torque on the base was also implemented. The functionality of each mode was demonstrated experimentally.
The work presented has provided the underlying theory necessary for generating low torque path that can be optimized using path planning and control for the modular serial-paired climbing robot.