Adaptive Cloud Offloading of Mobile Augmented Reality

A method comprising:

• receiving, by an augmented reality system, an acquired image frame captured by an acquisition device, wherein the acquired image frame indicates a plurality of acquired image frame parameters;
• determining, by the augmented reality system, a plurality of augmented reality process instances corresponding to different process modules that can be executed locally on the augmented reality system or can be executed remotely via an augmented reality cloud server to provide feature point extraction for the acquired image frame;
• determining, by the augmented reality system, a plurality of local feature extraction time estimates for the plurality of augmented reality process instances based on the plurality of acquired image frame parameters, wherein a local feature extraction time estimate is determined for each of the plurality of augmented reality process instances, wherein each of the plurality of local feature extraction time estimates provides an estimated amount of time for locally executing a respective augmented reality process instance on the augmented reality system to extract feature points from the acquired image frame;
• obtaining, by the augmented reality system, a network condition measurement;
• generating, by the augmented reality system, an offload command for each of the plurality of augmented reality process instances such that a plurality of offload commands are generated for the acquired image frame, wherein generating each of the plurality of offload commands is based on the network condition measurement and at least one of the plurality of local feature extraction time estimates;
• and
• selecting, by the augmented reality system, one of the plurality of offload commands to activate a respective augmented reality process instance.

A computer storage medium having computer-executable instructions stored thereon that, in response to execution by a processor, cause the processor to perform operations comprising:

• receiving an acquired image frame captured by an acquisition device, wherein the acquired image frame indicates a plurality of acquired image frame parameters;
• determining a plurality of augmented reality process instances corresponding to different process modules that can be executed locally on the augmented reality system or can be executed remotely via an augmented reality cloud server to provide feature point extraction for the acquired image frame;
• determining a plurality of local feature extraction time estimates for the plurality of augmented reality process instances based on the plurality of acquired image frame parameters, wherein a respective local feature extraction time estimate is determined for each of the plurality of augmented reality process instances, wherein each of the plurality of local feature extraction time estimates provides an estimated amount of time for locally executing a respective augmented reality process instance on the augmented reality system to extract feature points from the acquired image frame;
• obtaining a network condition measurement;
• generating an offload command for each of the plurality of augmented reality process instances such that a plurality of offload commands are generated for the acquired image frame, wherein generating each of the plurality of offload commands is based on the network condition measurement and at least one of the plurality of local feature extraction time estimates;
• and
• selecting one of the plurality of offload commands to activate a respective augmented reality process instance.