Wheeled mobile robots usually need wheel velocity measurement to implement the control system of the vehicle. For small-sized robots, usually used in robotics competitions, the assembly of encoders for velocity measurement of the wheels is not an easy task due to the space available to attach them to the wheels. Besides that, the computer vision system mounting over the arena of robotics competitions to measure the posture of the robot usually does not provide enough rate that allows an adequate response for the wheel velocity control system of the robot. Another option would be the use of an Inertial Measurement Unit (IMU) attached to the body of the robot, although integration of IMU data to obtain velocity is subject to drift problems. In this way, this project proposes an estimating system based on complementary filters to fuse computer vision posture and inertial measurement unit data. And to attenuate drift errors, detectors for zero velocity and zero acceleration will be used to eliminate numeric integration errors of the signals of the sensors. To evaluate the proposed approach, a simulation of the system using the robot CoppeliaSim software will be performed with the algorithm implemented in Python. And actual experiments will be performed with the wheeled mobile robots of the Red Dragons UFSCar Robotics Team. Simulation and actual results will be analyzed to evaluate the effectiveness of the proposed approach.
News published in Agência FAPESP Newsletter about the scholarship: