<-
The second issue is that there is no guarantee that Moore’s “law” will hold true 30+ years into the future. It’s not a real law, like the law of gravity. It is an observation of semiconducting processor performance by Intel co-founder Gordon Moore in the early years of semiconducting microprocessors [3]. There is mounting evidence this law is not holding true even now [4]. Given scrutiny, the suggestion that Moore’s observation is truly a law and would extend out to 300*10^12 transistors within a semiconductor-based microprocessor is not reasonable. Even Gordon Moore believes that semiconducting microprocessors are approaching the fundamental limitations of “the finite velocity of light and the atomic nature of materials” [5]. Certainly there are other computing technologies, both in existence and yet to be discovered, that may one day approach or surpass the complexity of the human brain. It is not, however, logical to use Moore’s law to support a conclusion that robots will have brains based on semiconducting microprocessors with the complexity of the human brain.
Automatic Pedestrian Door Example - At first glance, it is not obvious why an automatic pedestrian door example application would be appropriate for a paper on layered robotic system safety and collaborative robotics. Recall the earlier definition of a robotic system as, “a system with moving parts, a drive mechanism to drive the moving parts and a computer to choreograph and control the system.” With this understanding of a robotic system, it can be seen that an automatic pedestrian door is a robotic system with at least one degree of freedom. We have decades of experience with people operating in the same workspace as automatic doors.
Collaborative robotics experts could learn much from automatic door
experts.
Early automatic door systems employed control mats. The control mats were connected to electrical contacts that opened the doors when there was weight on the mat and closed the doors when there was no weight on the mat. There was very limited computing involved and hence these early automatic doors did not fit our definition of a robotic system very well.
Later automatic door systems began replacing the control mats with non-contact sensors, typically mounted above the doors, to detect people. To analyze the signals from these sensors and determine whether (or not) to open or close the doors based on this sensor feedback the automatic door systems include computing elements. One of the sensor types that found wide deployment was a motion sensor based on the Doppler frequency shift. This type of motion sensor identifies the approach of people towards the doors. The computing element then commands the doors to open for the people to move through the door way. Unfortunately, there were numerous accidents where some people moved so slowly through the doorway that they effectively became invisible to the motion sensing. This danger is exacerbated when the person is directly under the sensor because there will be little or no component of that person’s velocity towards or away from the motion sensor.
To address this issue, the designers of the automatic door
experts added an infrared presence sensor intended to identify the presence, rather than the motion, of people that could get into the path of the moving door. This is an important lesson to the designers of safety systems. Thought should be given during the design process to possible future safety requirements that may evolve in response to lessons learned and changing safety standards. Supporting the ability to add additional safety sensors is a reasonable design goal, but the control system must also allow these new sensors to be added in a failsafe fashion. That is, the control system must be able to identify very quickly if the new sensor has failed or become disconnected and respond in an appropriate fashion.
→
|
