Neshant wrote:
The tracking is in the post-processing stage, not necessarily real time - at least for now.
I dont' think its that difficult especially with some pre-programmed human input on what to look for and areas of interests outlined in the scene.
You are back to square 1 and my point. If you knew the area of interest, you would just use the present capability of focusing at that area and then moving from there.
Neshant wrote:
You look for movement and narrow it down from there. In the case of India, we are talking about basic capabilities first - because even the basic persistent stare capability is non-existant.
There is no use of "basic" persistent stare capability. You would not know what to do with that amount of videos. Even with current UAV systems, the bottleneck is how do you analyze the continuous stream of data coming from the UAV.
Persistent stare is useful in cases (as you said) where you are investigating an event (and hence know your point of interest). It can also be useful, if there is processing which can identify points of interests and then human beings can look at those points of interests. There is also a requirement of analyzing intelligent ways of discarding useless data. Otherwise, you won't be able to store data worth more than 1 week old.
Neshant wrote:
You have an auto-pilot with inertial navigation or GPS, a camera and what else? Somehow that stuff sounds low tech. Its what hobbyists are doing.
Persistent stare is really something challenging with India should pursue. There is a pressing need for it both for law & order and for other civilian purposes. I really hope Indian UAV program takes on the challenge of developing it. Developing UAVs that fly with a camera with a network link - that's old hat now. Time to up the challenges.
For a hobbyist UAV like mine it is fairly simple. There are of course some predefined flight profiles, like return to base etc. But for a full blown MALE, HALE, it is very very different. The first thing, is to have something which has a MTOW of 24 hours. This is easier said than done.
1. It has to be a full blown plane which can take such long endurance flights and survive 30-40 years. not many planes of that size have 24 hour endurances.
2. Now because it is a full blown plane, in case of Rustom-2, you have to make sure that it doesn't collide with another plane or drop on to civilian areas. I did not think of this one till I watched a video at AI'13 about the certification of such planes.
3. Also auto landing and take off of such full blown planes are very different and complex. If you have seen a model UAV land, it's flight path is quite generously. This is not the case a full blown plane ( ascale model of Rustom-1 was lost due to this). When Airbus (the pioneer) tried to implement auto-pilot during landing plane, they faced a lot of problems too. Their demonstration resulted in catastrophic crash.
4. For hobbyist, all the servos are electric and as long as you have juice in your battery, you will be fine. For a military UAV of over 2 tons, it is all hydraulics. To develop hydraulics which can work from -55C to 50C, in rain, salt and dirt is not an easy task.
5. The there is communication. You have to maintain secure and uninterrupted communication with the UAV, even when it is not in NLOS. For Low altitude drones, NLOS breaks of at 300 km. Then you have to work with relays. Now you have relays in serial which decreases your MTOW.
6. Latency, when communicating over large distances, the latency introduced by all the relays makes the plane and the operator out of sync. This is a major problem for global operations. Only US has a solution to this as of now.
7. And then of course there is the issue of payload components. And no they are not just stabilized cameras.
Think about it, if making MALEs and HALEs were so easy, how come the world has a handful of them. Whereas in USA, every town has at least one hobbyist who has a toy UAV. India needs to get Rustom-2 right. And ADE is making very strong steps towards it. They are going to start testing it in about a year from now. From a tender of work they just released they are going to be done with testing 3.5-4 years from now.