To control robot-driven machines, we must know their status, such as the position and orientation of the platform. The status is usually measured by conventional sensors, that have various limitations like a high cost, a difficult maintenance, and demanding deployment.
In his research, Zhang chose the Inertial Measurement Units (IMUs) as the main sensors because they could work independently without external signals and have a high output frequency. The biggest problem is that IMUs have bias, and the error of estimation will continue to accumulate. Fortunately, the Earth's gravitational field can be used as a reference to partially correct for these deviations.
In Zhang’s research, a heavy-duty forest machine was used as a platform. First, the estimation algorithm of the state of the robotic arm was tested on a floating platform.
“This platform on which the multi-joint robotic arm is installed is not fixed, and the attitude and position of the platform can be continuously changed. It brings some challenges to the development of algorithms. Many traditional algorithms are mainly aimed at the state of the robotic arm on a fixed platform,” Zhang says.
Zhang used low-cost IMUs, the devices attached to the surface of the robotic arms to successfully estimate the angle of the joints, while the heavy-duty machine was moving on gravel.
“One of the objects for the solutions is that the algorithms do not dependent on external signal, such as GPS. IMU can work reliable for a while without other sensors’ help once it is calibrated well. Because such machines often work in forests or underground, where satellite signals are invisible, it is difficult or costly to deploy other signal sources," Zhang remarks.
He also developed an algorithm for estimating wheel rotation using IMU. The traditional theory of wheel speedometer often perceives problems when the wheel speed is very low. If we want to accurately control the vehicle in a harsh and complex environment, we must accurately know the rotation state of each wheel.
The tests were done on a snow-covered field. At this time, IMU was used to overcome the problem of inaccurate wheel speedometer on a slippery ground.
The doctoral dissertation of M.Sc. Xiaolong Zhang in the field of engineering titled Towards IMU-based Full-body Motion Estimation of Rough Terrain Mobile Manipulators will be publicly examined in the Faculty of Engineering and Natural Sciences at Tampere University, Hervanta campus, in the auditorium Pieni Sali 1 of the Festia building on 16 December 2022, at 12 o’clock. The Opponents will be Professor Arto Visala from Aalto University and Program Manager Arto Peltomaa from DIMECC Oy. The Custos will be Professor Jouni Mattila from Tampere University.