C4 Cloudlets Passing in the Night

How do Cloudlets work in C⁴, pronounced “See Forth” – Cloudlet Communications, Computing & Control?

Let’s follow an Autonomous Vehicle through a scenario and see how it operates and reacts to other Cloudlets.

Let’s start with our Autonomous Vehicle out of range of local communications with any other Autonomous Objects. The Autonomous Vehicle forms a Cloudlet of its own in this situation, doing the same sorts of things it would in a larger Cloudlet, but not having any other Autonomous Objects to communicate with inside a Cloudlet. Its information is limited to what its sensors can detect, plus the static maps it has acquired, and possibly dynamic local information from Autonomous Objects previously traversing this local area (the date and time stamp will give an indication of how out of date this dynamic local information is, recall there aren’t any other Autonomous Objects currently in the local area – I’ll discuss this further later in the scenario).

Functioning as a Cloudlet, this Autonomous Vehicle creates and stores a current 4-Dimension Local Dynamic Map based on the information it has available. It sends out “Cloudlet here” signals and communicates with higher levels in the communications hierarchy, as a Cloudlet.  For example, if it finds discrepancies between the local map it creates and the static information it enters that into the management protocols – note, because there is only a single Autonomous Object in the Cloudlet, this information will be treated as less reliable than if it were from a larger Cloudlet, which would have included multiple points-of-view and extensive internal validation before making such a report, for example to account for variation in sensors, because accuracy and redundancy are key features of the C⁴ system due to the critical uses of the information.

Now suppose this Autonomous Vehicle detects another Cloudlet. The first step is to establish communications Cloudlet-to-Cloudlet. The Cloudlets exchange significant information about travel conditions (e.g., slippery road, accident ahead, potholes, etc.). The Cloudlets determine their relationship to each other, for example, should they merge or not.

Let’s assume for this scenario that they will not merge. There are many possible reasons for this: they could be on separate roads (e.g., one is on an over-pass, or going in a different direction on a divided road; note if this were an A-Way, the A-Way would likely block local transmissions between separate A-Ways, and pass through any relevant information).

Assume our Autonomous Vehicle is approaching a Cloudlet going in the opposite direction on an undivided road. This approaching Cloudlet is a cluster of Autonomous Vehicles. Because the Cloudlets are passing each other rapidly, and will not have comparable points-of-view, they decide not to merge into a single Cloudlet, but to continue to function Cloudlet-to-Cloudlet.

One important set of information to share is the Physical Boundary of the Cloudlets (unless stated otherwise, this always includes 3-dimensional information on position, velocity, and acceleration). For example, is one of the Autonomous Vehicles in the Cloudlet passing others in the approaching Cloudlet, and thus in the lane occupied by our intrepid Autonomous Vehicle? In this case, the passing vehicle would have it’s Plan in place, showing when it will complete the passing maneuver, and thus our Autonomous Vehicle can determine whether there is danger of a collision or not. Or, is the Cloudlet performing normally, for example if one of the vehicles were controlled by an intoxicated driver, it might be weaving and even crossing the lane boundary, and cause our Autonomous Vehicle to report this anomalous behavior using the Management Protocol, and potentially to take evasive action depending on distance, speed, acceleration – hopefully we don’t have to worry about intoxicated Autonomous Vehicles, because that’s one of the major advantages of Autonomous Vehicles over human drivers (over 90% of crashes are due to human error). J

The corresponding Physical Boundary information to the other Cloudlet about our Cloudlet would indicate that their Autonomous Vehicles should not pass, and need to maintain suitable clearance. For example, our Autonomous Vehicle might be carrying a wide load, and thus other Autonomous Vehicles should give it additional space.

In sharing information on maps, they are coded to indicate whether they are the most recent version. Today as an airplane is approaching an airport, it gathers the latest weather and other airport information, and when opening communication with the tower, it first gives the code of that information, e.g., ZYING, so the tower knows it has the latest information. Thus if the approaching Cloudlet has detected a change in the static map information, it will be immediately clear and our Cloudlet can request the new information.

Note we do not need to know information about the individual components of other Cloudlets, and vice versa, that’s one of the key differences between a Cloudlet and an Autonomous Vehicle, even if the Cloudlet only contains one Autonomous Vehicle. This has implications for privacy, which I will discuss later (I know I’m putting off a lot until later, but life is complex).

Our Autonomous Vehicle, and thus Cloudlet, has indicated our travel plan. For example we may be turning right at the next intersection. This helps determine how much information in the other Cloudlet’s Local 4-DMap it needs to share. In this case, most of the information beyond the next intersection is irrelevant, although it would include information on Cloudlets that may reach the intersection before we do, and thus have relevant information.

If there are discrepancies between the information from the two Cloudlets, then management protocols exist to help discover the cause of the discrepancy and to resolve it; or to report it to higher levels in the management protocol if they cannot be resolved. For example our Cloudlet might sense that there is something ahead, but the other Cloudlet does not sense it. In this case because the two Cloudlets are separate, there is not as much opportunity for overlapping points-of-view, mutual calibration of sensors, etc. so it may just be that the offending object is blocked from the other Cloudlet’s Point-of-View. I’m reminded of the scene in Goldfinger, where James Bond is driving down an alley and sees headlights approaching; he fires his Aston Martin’s machine guns, but to no effect; at the last minute he sees that he is approaching a mirror – sensors don’t always give the full picture.

As our Cloudlet moves past the other Cloudlet, they move out of local communication range, or otherwise out of mutual relevance, and they disengage. Each Cloudlet has the latest version of the consolidated dynamic local 4-DMap, and they proceed on their own ways.