The Relational Data Format of the traffic grid by Navteq, the various XML map definitions used on the web, and the Wave protocol for traffic messaging between traffic devices. What are the essential ingredients? What is the object, where is it, and where is it going. The Robocar knowledge of truth is a list of objects in its final map, the list describing the objects, where they are and where they are going. The entire robocar traffic network , including traffic signs follow this pattern. A stop sign sits where it is and isn;t going anywhere, but a car may proceed through the intersection; the description format applies in both cases, even the traffic rules for the DMV use that format.
The overall langague of traffic is simple, because the only applicatipon the language really supports is computing the next configuration of objects. Autopbots are simple, they take the known array of objects, compute their own next position, go there and repeat, objects in traffic simply rearranging themselves.
What is the math behind it all?
Well it is a predictive fileter, taking the object map at time t and computing an expected map at time t+1, including the objects own expected position. Moving to a new position in the map requires taling the approopriat rule section of the rule book and overlaying that on the map to compute exclusion and inclusion zones. But the zones themselves are staionary objects with positon, nothing more. The math behind the predictive filter carrys both the predicted position of an object and an error term defining how much error is in that position. The predicted position and the error term being updated whenever a measuremnt on the object arrives. Generalized measurments of objects follow the same format. These come from smart cameras, laser radar, GPS, communications, or digital mars.
The outcome of all this real time mapping are the control signals to steering, brake and acceleator, as well as continual communication with other traffic objects. So, other than a simple dashboard display, there is no real human involved.The key to this industry is understanding how the autbots can simply arrive a common conclusions without human interaction.
- A common definition of traffic rules.
- A shared description of vehicle and object types.
- Common messaging between smart sensors and autobot navigators.
What does it take to get this going? It is happening,and fast. Vehicle designers beware, control by wire steering, braking, and acceleration are a must have. The rest can be added later with a black box upgrade. Ray LaHood has no choice but to support higher vehicle speeds and larger vehicle configurations on HOT/BRT lanes with road pricing; using robocar technology.
No comments:
Post a Comment