Predictive Situation Awareness – the future for mobile workforces?

Posted by Ross Coundon on September 25, 2012

Recently I was honoured to be invited to deliver the keynote speech at the Ryzex technology day. The topic I was asked to speak about was “Big Trends in Mobile” and to make things interesting I thought I’d give the audience a light-hearted spin through the history of mobile technology. I then spoke about some current trends, mobile strategies and finished with some predictions about where I think we might be going with enterprise mobility and mobile working.

I thought it useful to write about one of those predictions here – Predictive Situation Awareness.

So let’s start with Situation Awareness. Wikipedia defines it as:

Situational awareness is the perception of environmental elements with respect to time and/or space, the comprehension of their meaning, and the projection of their status after some variable has changed, such as time. It is also a field of study concerned with perception of the environment critical to decision-makers in complex, dynamic areas from aviation, air traffic control, power plant operations, military command and control, and emergency services such as fire fighting and policing; to more ordinary but nevertheless complex tasks such as driving an automobile or bicycle.

Whilst Situation Awareness provides some ability to determine future events and to adapt the operation proactively, it’s principally concerned with the now and how to best deal with complex conditions as they unfold.

For a field-based organisation these complex conditions might include:

  • Engineer locations
  • Engineer workloads
  • Engineer skill levels
  • Van inventory levels
  • Contractor locations
  • Asset locations
  • Asset history
  • Customer reported incidents
  • Detected incidents
  • Weather conditions
  • Traffic
  • Roadworks

Producing a coherent picture of the operation with all of these variables is extremely difficult. As more and more data is received into the operation from machine-to-machine probes, mobile and other data feeds it becomes increasingly difficult to turn this data into useful information that can support critical decision making.

This is where automated scheduling and optimisation tools can help make sense of this torrent of data. These tools are getting more sophisticated all the time and can make sense of far more variables than a human mind. The basic goal of these optimisation engines is to solve the Travelling Salesman Problem ( as efficiently as possible but also take into account many of the necessary variables listed above.

During normal operation, it is the machine that is aware of the situation. In those scenarios where it is not possible for the engine to make a scheduling decision, an exception is presented to a human dispatcher. For example, a judgement call may need to be taken between sending an engineer to fix a fault affecting two customers vs a fault affecting a single VIP customer. There are some kinds of decision you just don’t want to trust to a computer.

These kinds of technologies can offer vast improvements in efficiency, productivity and customer service levels through the ability to do more with less, improve first-time-fix rates and offer more granular appointment times.

In the future, I see this moving to another level entirely. As optimisation engines become more sophisticated I believe it’s only a matter of time before Artificial Intelligence is used to appraise situations and predict future events. In fact, a number of years ago I worked on a project that dipped its toe into these waters. A database of scenarios was created that meant that when a probe in the customer’s network detected a fault, this could be looked up against a past history of fault, cause and reason codes. The output of this was a probability of what exactly the fault might be, the skill level of the engineer that must attend the site, what is required to fix it and the level of access required in order to get to the equipment in question. The system didn’t learn by itself, it needed human input to create the scenarios but the principle is the same. This meant that the operational efficiency was improved drastically since the diagnosis could be done statistically prior to attending site therefore saving time and effort.

What if, instead, the system could identify and learn from these scenarios? Here the machine would be continually appraising all of the data feeds, looking for high probability failure events and reshaping the operation automatically and proactively to deal with them before they occur.

To take an example for a water utility company, the system could look at predicted weather data, water pressure across the network from telematic probes, reported leaks, previous choke/blockages in the area and location of engineers. With this data it could predict that at 1:30pm when the rain in Basingstoke begins there is an 80% likelihood of a minor flooding event. Using this prediction it then begins to move skilled engineers into the area in order to either react more quickly should it occur or actually perform some preventative maintenance to stop it occurring altogether.

This technology might be a few years away commercially and I’d be surprised if the military isn’t looking into this for battlefield intelligence since the principle is the same. When it does arrive, I believe it will transform the way we look at field operations.