Prototyping a Fault-tolerant CAN Bus Based Distributed Servosystem

Distributed Realtime Process Control Systems
Project Period:
PR7, 4. September - 13. December, 2002
Introduction: In conventional car power steerings the steering arms on the hinged wheels are mechanically connected by a track rod. This construction has the advantage of being simple and cheap. It is, however, quite space consuming and vibrations from the wheels are transmitted through the steering column to the steering wheel. This raises the idea of a distributed servosystem, utilizing electric actuators to position the wheels and for force feedback to the steering wheel. The overall functionality of such a system must correspond to that of a conventional steering gear.

Methods and Material: A model of a power steering has been constructed. It consists of a modified off-the-shelf PC force feedback steering wheel and two electric servomotors, each representing a wheel actuator. Each actuator is attached to a network node based on two Peripheral Interface Controllers (PICs) (PIC18F458) used in a master/slave configuration. The two PICs form a redundant system, intercommunicating via a Serial Peripheral Interface and with either PIC connected to one of two CAN lines. This is done to preserve the network communication between the network nodes even in case of a cable breakdown.

Since only one PIC can control the actuator and no supervisory arbiting circuit has been implemented, the PICs themselves must decide which should have the motor control rights. The PICs supervise each other through data synchronization and "handshake" at certain breakpoints in the program. If one PIC fails to respond in a handshake the actuator control is handed over to the other.

Results: The test strategy was to invoke a number of faults one by one (in accordance with the Single Point of Failure approach) and thus cause the system to fail.

The tests gave satisfactory results and showed that the functionality of the system's fault handling and redundacy work. On the other hand the reaction of the wheel actuators has turned out to be slow.

Discussion: The main area of interest in this project was the distributed system. Therefore, only limited work has been done on motor control and motor modeling. The result is that the system does not act as precise as a conventional steering gear due to the use of too weak actuators.

Further work is needed in the alarm signaling and handling. The focus has been on the serious errors meaning that the status of the system will not be known until the occurence of one of these.

Group members:
Michael Skipper Andersen
Jørgen Friis
Niels Nørregård Hansen
Johnny Jensen
Rene Just Nielsen
Michael Pedersen
Roozbeh Izadi-Zamanabadi
5 pages
96 pages

Download article
Download worksheets
Download poster