Custom Search

Wednesday, January 7, 2009

Sensor Algorithm and ULP WSN Go Hand-in-Hand

In recent months, there has been a very emotional debate in the Continua Health Alliance technical working group regarding the inclusion of broadcasting as a mechanism for sensor data transmission. In the current (version 1) Continua standard, only the point to point (P2P) wireless link is specified (Bluetooth). The session layers above stipulated by IEEE11073 are also based on P2P connections. This is understandable as it best address the needs of many current medical sensor use cases where reliability are strictly required. Now in the process of specify version 2 standard, it is pointed out that P2P solution is power hungry, with large overhead and is not energy efficient to address the ultra low power requirements of many wearable sensors. Dynastream, the inventor of ANT, is the leader lobbying Continua members to accept broadcasting. In Continua Boston summit, a presentation was made by Dynastream, and got support from some other companies like Polar who are now offering ultra low power sensors today. But the resistance is also high. It seems a lot of concerns are related to the relationship with version 1 standard and are from Bluetooth camp.

The debate is in truce now and will resume soon after new year. I went through Dynastream's presentation again and tried to peel off the political shell to view the argument from a neutral standpoint. I see this presentation from an ultra low power wireless sensor practitioner is really thoughts provoking. Many points in the presentation can be further explored. But here I just want to share the most critical in my view, the development direction and trend of the ultra low power and wearable sensors.

In Dynastream's description, using broadcasting, unlike point to point communication, a sensor just broadcast the data and receivers selectively receives data (similar to cable TV). Sensors basically do not care the receiving party nor the successful delivery of data. At first glance, broadcasting seems a technology inferior to P2P which consists of complex process like handshaking, authentication, acknowledgment and retransmission etc. But as Dynastream claimed there is a type of sensors that really do not care reliability as much as other sensors, and because of this trait, broadcasting can be used to send information efficiently. Trade-off between power saving and reliability is doable and has been proven by market. Dynastream claims with the acquiescence that all the current ultra low power sensors in sales for the sports market today are using broadcasting. Such type of sensor has the following characteristics:

1. The data to be transmitted wirelessly is changing slowly (per second)
2. So missing one or two data is not critical
3. And the latest data is more important than those missed

Examples of this type of sensors include body temperature, heart rate, weight etc. In comparison to this kind of sensors, those fit to use P2P communication, for example, ECG, blood pressure, need to transmit reliably every bit of data, so that a central computer unable to put on body can process the data and deduct useful information. In the presentation, Dynastream used "complementary" to plot the relationship of P2P and broadcasting to address the two different kind of sensors.

Behind broadcasting is an entire different philosophy!

The data transmitted from the ECG leads or Blood pressure cuff have to be received accurately and reliably, but the information deducted, e.g. heart rate, heart beat intermittence, blood pressure are quite slow changing. Thus, if the computation function can be moved to the leads or to the cuff, then broadcasting can be used without much sacrificing of reliability. So to make a wearable ultra low power wireless sensor is all about to move the computation function from a central hub to the sensor node.

Let's explore this further by examining one of the popular use sensors, foot pod. There are two types of foot pods, both measuring a runner's speed and distance. Nike+ foot pod uses piezoelectric sensor and counts the time intermittence between two steps. From there, an algorithm deduct the running speed and distance. ANT+ foot pod is more accurate and complex, which uses accelerometers to measure the movement of your foot to deduct the speed and distance. No matter how a foot pod works, it does not transmit the raw data, rather it calculate the data to obtain the information of interest and then send or to be accurate, broadcast it using ultra low power wireless technology. Should the raw data has to be sent, then the P2P connection has to be used to ensure reliable delivery of every bit of the data or "waveform", so that useful information can be calculated by a more sophisticated computer.

In summary, if a sensor is to measure a direct input, like temperature, weight, heart rate, ON/OFF etc, it fits to use ultra low power wireless; likely broadcasting is the most efficient way. If a sensor is to measure something deducted from some direct inputs, depending on the size and complexity of the algorithm. If the algorithm is compact enough to be able to fit in a small memory size and resource and power limited MCU, then the sensor fits ultra low power. Otherwise, P2P connection has to be used. In the last case, the sensor can hardly be 24/7 wearable with long battery life.

So besides the wireless technology, another R&D area of wireless sensor network is the algorithm of all kinds of sensors that enable distributed computation to allow useful information deducted right at the sensor node.

Algorithm and Ultra Low Power wireless sensor network go hand-in-hand.

No comments:

Post a Comment