We have developed a tactile sensor network system, which consists of integrated MEMS-LSI tactile sensors (sensor nodes), relay nodes, a host and flexible bus power/interconnection lines for mounting a large amount of tactile sensor elements on the whole body of robots. Each sensor node, which is connected on the common bus lines, transmits sensing data independently to the relay node on the bus when exceeding a predefined sensing threshold value. There are system-level issues such as a threshold value choice for sensing data transmission and a throughput control considering data collision on the bus line. Before fabricating the integrated MEMS-LSI sensor devices, performance of the whole system should be estimated, because to fabricate the integrate devices is expensive. In this paper, we report a less costly simulated network system for resolving the issues without the integrated devices. The simulated system uses easily-available components such as discreet sensors, FPGAs (Field Programmable Gate Array) and a PC. In the system, we implemented sensor node monitoring circuits to analyze network characteristics precisely. The circuits cannot be implemented in the real integrated devices and systems. By using this simulated system, we analyzed data collisions and average transmission delay time. The experimental results show that in the case of 20 sensors pressed with strong forces simultaneously the data collision rate is about 44% when the transmission rate is 1Mbps and about 0.7% when the transmission rate is 10Mbps, respectively. We found that when more sensor nodes transmit their data on the bus simultaneously and/or the transmission rate is lower than several Mbps, each sensor node should control the network traffic for avoiding the collisions with a more smart control mechanism than a simple mechanism we applied.
抄録全体を表示