The REIMEI satellite launched from the Baikonur Space Center in August 2005 is a small scientific satellite weighing about 70kg. The satellite performs aurora observations using onboard camera and various sensors (Fig. 1). An attitude control function to point the satellite in a target direction is required to carry out the observations. On REIMEI, the function is implemented mainly by an electrical magnet called a “magnetic torquer,” which was introduced to reduce the satellite weight. This method uses the faint torque that arises between the electrical magnet and the earth’s magnetic field. We developed a technique to estimate the slight magnetism that the satellite takes on, which is detected by sensing fluctuation in the satellite attitude. In this way, we achieved high attitude control performance of about 0.1 deg. error – an outstanding level for a small satellite.
One of the concepts for the REIMEI satellite was “in-house development.” For the attitude control system, this was realized by the approach that, although hardware instruments such as sensors were sourced from outside, tests. for the instruments were carried out by us (in-house). In addition, all the software programs to integrate the instruments were written by the in-house development team. Members of the team responsible for the development were amateurs. Moreover, none of the team, including myself, who had just started work at ISAS, and other students had any experience of satellite development until that time. Naturally, the development was a process of trial and error. In this article, I would like to discuss our experiences of troubles, new ideas, etc., in the development of the attitude control system of the REIMEI satellite.
Design of the attitude control system utilizing a simulator
When I was at university, I was taught, “Never design a control system without tuning by experiment.” With satellites, however, we never experiment until the launch. Before REIMEI, I had no connection with the space field. At first, I had problems with “development without experiment.” My last resort was to use computer simulation, so I planned the following tactics.
The technique used for (1) is called the Monte Carlo method in which parameters are generated according to random sampling numbers. This method takes immense computation time because, for example, for a computation requiring 20 minutes per case, we have to execute each one after changing the conditions. In fact, our record shows that more than 40,000 cases were computed.
To perform the computation, we built a distributed-processing system where multiple PCs are linked via a network. It is not large system, and its construction took only one week. We created an automatic program that distributes initial-condition files via the network to start computation and collects the results after computation. With the cooperation of the development team members, the system was installed onto about 10 PCs.
This system helped us to avoid problems. When the REIMEI launch rocket was changed, we simulated about 2,000 scenarios just in case. As a result, we found that the sun acquisition by the satellite performed incorrectly in several cases. With careful examination, it was revealed that there were certain casewhere the relationship between the earth’s magnetic field and the direction of the Sun falls into a very particular, fateful condition for a new orbit. We immediately corrected the problem by revising the control laws. The problem was discovered about six months before the launch. I am appalled to imagine what would have happened were it not for a system capable of performing a number of simulations in a short time.