Adaptive scheduling of streaming video over wireless networks
View Patent ↗An adaptive scheduling process is disclosed which dynamically decides which frames need to be transmitted and which ones need to be dropped at any transmission opportunity based on current channel conditions and on characteristics of each video frame.
1. A method to transmit streaming video over a wireless network channel with periodic transmission opportunities, comprising:
receiving a plurality of video frames for transmission, each frame having an importance θ and a deadline by which the frame is to be received;
transmitting video frames at each transmission opportunity with a scheduler, wherein the scheduler maximizes the total importance for all frames to be received within each frame deadline by applying dynamic programming on a predetermined window of frames and updating the scheduler with a channel feedback;
applying random channel error probability in selecting the frames;
determining an expected utility u(i, T) as attainable by deadline T with up to i frames; and
recursively determining the expected utility u(i, T) as u(i, T)=max (u(i−1, T), max (u(i−1, T−k)P li,k ·θ i , over l i ≦k≦T and P li,k ≦1−δ)) where
P
l
,
k
=
∑
j
=
l
k
(
k
j
)
(
1
p
e
)
j
p
e
-
k
-
j
is the probability of having at least l successful transmissions out of k attempts, l i is the length of time of frame i, and p e is the packet error probability at one transmission.
2. The method of claim 1 , comprising determining u(i, T) as a maximum utility (importance) that can be attained with a deadline T and with frames up to i.
3. The method of claim 2 , comprising determining u(i, T)=u(i, T i ) if T is greater than the deadline T i of frame i and otherwise determining u(i, T) first recursively and then using dynamic programming.
4. A method to transmit streaming video having a plurality of frames over a wireless channel, each frame having an importance measure relating to video quality, comprising:
maximizing a total importance measure for all frames to be received within each frame deadline through dynamic programming with a predetermined window of frames;
dynamically scheduling one or more frames to be transmitted or dropped at each transmission opportunity based on current channel conditions and on characteristics of each video frame;
updating the dynamic schedule with a channel feedback;
applying random channel error probability in selecting the frames;
determining an expected utility u(i, T) as attainable by deadline T with up to i frames; and
recursively determining the expected utility u(i, T) as u(i, T)=max (u(i−1, T), max (u(i−1, T−k)P li,k ·θ i , over l i ≦k≦T and P li,k ≦1−δ)) where
P
l
,
k
=
∑
j
-
1
k
(
j
k
)
(
1
-
p
e
)
j
p
e
k
-
j
is the probability of having at least l successful transmissions out of k attempts, l i is the length of time of frame i, and p e is the packet error probability at one transmission.
5. The method of claim 4 , comprising deciding frame(s) to be dropped based on deadline, size and importance of each frame.