1994 Fiscal Year Final Research Report Summary
Autonomous Decentralized Control of a Mechatronics System Composed of Intelligent Actuators
| Project/Area Number |
05650381
|
|---|
| Research Category |
Grant-in-Aid for General Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
計測・制御工学
|
|---|
| Research Institution | Tokyo Institute of Technology |
|---|
Principal Investigator |
MATSUO Yoshiki Tokyo Institute of Technology, Faculty of Engineering, Associate Professor, 工学部, 助教授 (90173806)
|
|---|
| Project Period (FY) |
1993 – 1994
|
| Keywords | Autonomous Decentralized / Actuator / Cooperative Control / Impedance Control / Disturbance Observer / 2 Dof Robust Control / Genetic Algorithm / Genetic Programming |
|---|
| Research Abstract |
As a fundamental study on distributed autonomous mechatronics systems, this research investigated autonomous decentralized control of a mechatronics system composed of intelligent actuators. The target system is a self-standing Caterpillar-like machine which comprises plural hinge-like subsystems each driven by a DC electric motor and controlled by a microprocessor. The task of the system is to roll across the floor only by autonomous decentralized control of each subsystem. After derived exact mathematical model of the whole target system, the investigators developed a simulation system employing 3-dimensional animation. Also, they constructd an experimental system composed of 8 subsystems.
As for the control scheme, distributed torque generating laws were studied at first. A torque generating law constructed as a function tree on various state variables was refined by a genetic programing method on simulations. The result clarified that the sufficient information for each subsustem to
… More
generate the torque reference are the grounding status, the relative angle and the relative angular velocity of the subsystem itself. Then, the torque generating law was reformulated as a numerical map on these three state variables and sub-optimized by a genetic algorithm on simulations. According to experiments, the sub-optimal torque generating law performs the rolling task and shows almost similar properties to that in the simulation, however it seemed rather sensitive to torque disturbances than expected.
Thus, another scheme was studied where each subsystem is controlled by a 2 degree of freedom robust relative angle controller. Then the response of each subsystem to the external torque is adjusted by a disturbance observer and a response model. In this control scheme, the whole system verified to be able to roll smoothly only by applying a proper size of step-shaped reference to the relative angle of each subsystem. Moreover, the same control laws worked well even if the number of subsystem was increased or decreased, or in the situation where a subsystem could not produce torque at all. Less
|
