Path planning for a UAV tracking a ground object and using an inertially stabilized camera platform


Our team has developed several prototypes of inertially stabilized platforms carrying optoelectrical payload such as day and/or night vision cameras, laser range finders and laser  markers. The platforms are intended to be mounted on mobile vehicles, in particular unmanned aerial vehicles (UAV). Together with implementation of HW and SW for these platforms, our group has developed and documented/published a control theoretic framework for this class of control systems. The platforms have been extensively tested indoor but also outdoor. Hence, a certain maturity can be claimed in the mere task of inertial stabilization and visual tracking.

The challenge that we are facing now is to extend the current scheme with the functionality of path planning. The current assignment coming from a defense industry partner is that in order to track the ground object, the only actuators are the motors driving the gimbals. No commands to the autopilor of the UAV were allowed. The path of the UAV was set by some algorithm or human operator. The task for the student involved in this project will be to break this constraint and formulate the problem of planning a path for a UAV in order to guarantee best visual tracking (and inertial stabilization). One and several ground objects will be considered. Moving and fixed.

The ultimate goal is not to come up with a fully autonomous solution. Such a system will not be realistic in the next decade or so for legal reasons. But it may be perfectly realistic and actually quite useful if some optimization problem can be formulated and solved in real time which assist the human operator in planning the right path, which is currently done manually and heuristically. Using numerical real-time optimization, it will be possible to include whatever restrictions on distances to the observed objects, UAV turning radii, but even the UAV wheels constraining the field of view. This way the work has a good chance to be included in a real system.

The interested student should have some inclination towards numerical optimization since the work will revolve around it. But no deep research knowledge of the subject is needed. Being an aerospace enthusiast might make the work more enjoyable too although only basic concepts from aerospace will be needed such as flight angle, roll, pitch and yaw rates and so on.

Contact person: 
Zdeněk Hurák
Contact person: 
Martin Řezáč
Image: Mi-17 helicopter carrying the inertially stabilized camera platform S250
Image: Inertially stabilized camera platform S250