SkyShare - Solution to Excess Check In Baggage
- 0 Collaborators
The headway of platforms concerning one more aspects of shared economy has received a propulsion of multiple dimensions since the turn of the past decade. Concerns surrounding security, authenticity and fraud management have transcended from limiting aspects to enabling ones – thanks to the cutting edge technological advances. A tricky utilitarian commodity that stands at the invigorating frontier of making a transition into the shared economy framework is the checked-in boot space on aircrafts. Been that person who’s had to pay for the excess luggage? We all have! And that’s a pain point we’re trying to address. We propose the development of a novel platform that would enable flyers of domestic aircrafts to share luggage space with co–passengers. The idea of allowing passengers with excess baggage to hand over some or all of it, to those with less than the allowed limit is both intriguing and challenging at once. A web based application will allow flyers (of a particular trip) to (a) Offer to carry others’ baggage (b) Request that their excess baggage be carried by co–flyers. An incentive based adoption model will drive the use cases, while a feature intensive estimation and matching algorithm will enable seamless user experience. ...learn more
Project status: Concept
Intel Technologies
AI DevCloud / Xeon,
Movidius NCS
Overview / Usage
Registered and verified users, once booked onto a flight can identify themselves in one of the two categories available for that flying session – a carrier or a seeker. Post the category selection, users are requested to furnish estimates of the excess luggage that can be associated with them. Based on luggage data, an estimation algorithm will fine tune the user provided estimates and associate with each of the users, a discrete Luggage Value (LV). 8 hours ahead of the scheduled departure, a matching algorithm will pick up the LVs of all participating users and begin mapping seekers with carriers. Iterative error accommodation and optimization will push category definite user notifications – with all required specifics. At the airport users will assemble at our assistance desk, have the baggage verified and transferred. A successful handover at the origin and retrieval at the destination will be logged by both the users and their records and incentives would be processed.
Methodology / Approach
A CNN based image processing algorithm augmented with linear combination of luggage content value, generates bounded LV components for each of the users. This simplifies the matching algorithm to distributing luggage with bounded LVs among all registered co-passengers, with none of them exceeding their own allowed limit. A negative cost (as a payment) is associated with the seeker, while a positive cost (as an incentive) is associated with the carrier. This optimization problem is solved using the 0-1 multiple knapsack algorithm to maximize the cost for each user, constrained with high self-luggage affinity of all carriers. A bipartite matching algorithm will handle these user affinities and pack a solid solution stack!
Technologies Used
We plan to develop the entire application based on Microservice Architecture, where each individual component feature will be separate service deployed on its own VM/Server. Therefore we will offload the entire training/inference service pipeline on Intel AI DevCloud during development, testing and production stage. For Developing CNN based LV Estimator we will use pretrained VGG Net at IntelAI DevCloud and retrain its last few layers to estimate dimensions of the Luggage. We can also run trials of Movidus NCS for our client site (airport) equipments for real time inferencing of LV of the users. Also for more mission critical system like LV Distribution we can utilise Intel Parallel Studio for developing 0-1 Multiple Knapsack algorithm which is a NP Complete Problem hence inherently require optimisation and approximation to be tractable.