Dr. Kashif Javed

The inchoate interpretation of the intelligent home assistant can be taken into account as an additional helping hand to carry out various daily chores. Though having rather manifold breakthrough aspects, we restrict to the prototype of a model that can markedly be rendered to rather notable extents. The preeminent principal of the assistant is to adhere to the orders of a specific person who throws words at the assistant. Adhering to the person�s will is further translated into the assistant performing the task, for example, grabbing a plate and traveling all the way back. The very central principal is rather a complex one to tackle, and constitutes whole three strides; the first one is to identify the voice of the speaker and interpret, as the training data set suggests, corresponding to an action revolving around an object previously a part of the training data as it would allow the assistant to recognize precisely. The second step includes equipping the assistant with a source of vision with which it can scan through the entire surroundings hoping to find the object around which the action circles. This has been done by assembling a camera with the assistant enabling it to view the object and identifying it as the center of action. The third stripe is to commute all the way up to the object and concluding the requested activity. The step feels smooth and simple to execute but an entirely opposite fact turns out to be true when we have to take into account of all the restrictions and hindrances on the way up to the object. Those hindrances can sure be avoided manipulating the obstacle avoiding mechanism into coping to our aid but this does not allay the problem that the trajectory is simply not stored in a memory to be re-traced back without any predicament. Currently the jeopardy is being tackled by introducing the Kaman filter that estimates the point at which the object is moving (in our case the robot is supposed to be moving possibly around an obstacle, the relative motion between the assistant and the object would serve the part). A cruder approach is to apply grids to the viewing screen and whenever the destined object goes out of the view it can be retraced back into view playing with the co-ordinates (for example the x-axis being positive or negative for leftwards and rightwards motion of the object). Once the trajectory has been concocted, the assistant commutes towards the object avoiding every hindrance and re-traces back to the initial point having dealt with the task assigned.

Download Link

Download Link