TOWARD OPTIMAL ALLOCATION OF LOCATION DEPENDENT TASKS IN CROWDSENSING Jingtao Yao Lab of Cyberspace Computing Shanghai Jiao Tong University

WHY I CHOOSE THIS PAPER  It is a paper for crowdsensing.  It is a paper published by InfoCom  It is a paper that studies the task allocation problem in crowdsensing.

ABSTRACT  Introduction  Problem Formulation  Prove Hardness  Local Ratio and LRBA  Theoretical Analysis  Pricing Sensing Task  Numerical Result

INTRODUCTION  Small-sized portable mobile devices are becoming extremely prevailing nowadays, that accelerates the emergence of crowdsensing applications.  Existing works:  Crowdsensing for specific sensing application  Unified platform  Incentive-based mechanism  This paper: Focus on location dependent task allocation

INTRODUCTION  This paper’s contributions are three folds.  Study the problem of allocating location dependent tasks and show that the formulated problem is NP-hard.  Design an efficient approximation algorithm, namely local ratio based algorithm(LRBA) to solve the proposed allocation problem and show that LRBA is a 5−approximate algorithm.  Design a pricing mechanism based on bargaining theory.

PROBLEM FORMULATION  Users u i, Task t j, Position P tj, Task set T i, Shortest Path P(T i )  Total time D(P(T i )), Time budget B i, times of Tasks l j  Reward R ij, Decision variable x ij

PROVE HARDNESS  One mobile users  G(V,E) cost for edge and rewards for node  Orienteering problem, NP-hard problem

LOCAL RATIO AND LRBA

 (A) Transforming the original MRP problem

LOCAL RATIO AND LRBA  (B) Solving the orienteering problem of each user forwards  f’ I−1 (y) to denote the reward function at the beginning of iteration I, and f I (y) the modified reward function at iteration I.  where  solve the orienteering problem associated with user u I  O I denote the assignment

LOCAL RATIO AND LRBA  that is  Iteration continued.

LOCAL RATIO AND LRBA  (C) Refining the final assignment backwards.  Note that in the second process, each sensing task may be allocated to multiple users at different iterations.

A EXAMPLE

THEORETICAL ANALYSIS

PRICING SENSING TASK  the probability that mobile user u i would accept the agreement

NUMERICAL RESULT

THANK FOR LISTENING ！