D B ] 20 O ct 2 01 5 Task Assignment on Multi-Skill Oriented Spatial Crowdsourcing ( Technical Report )
Peng Cheng,Xiang Lian,Lei Chen,Jinsong Han,Jizhong Zhao
2015-01-01
Abstract:With the rapid development of mobile devices and crowdsourcing platforms, the spatial crowdsourcing has at tracted much attention from the database community. Specifically, t he spatial crowdsourcing refers to sending location-based re quests to workers, based on their current positions. In this paper, we consider a spatial crowdsourcing scenario, in which each worker has a set of qualified skills, whereas each spatial tas k (e.g., repairing a house, decorating a room, and performing entertainment shows for a ceremony) is time-constrained, u nder the budget constraint, and required a set of skills. Under this scenario, we will study an important problem, namelymulti-skill spatial crowdsourcing (MS-SC), which finds an optimal workerand-task assignment strategy, such that skills between wor kers and tasks match with each other, and workers’ benefits are maximized under the budget constraint. We prove that the MSSC problem is NP-hard and intractable. Therefore, we propos e three effective heuristic approaches, including greedy, g-divideand-conquer and cost-model-based adaptive algorithms to g et worker-and-task assignments. Through extensive experime nts, we demonstrate the efficiency and effectiveness of our MS-SC processing approaches on both real and synthetic data sets. I. I NTRODUCTION With the popularity of GPS-equipped smart devices and wireless mobile networks [12], [16], nowadays people can easily identify and participate in some location-based tas ks that are close to their current positions, such as taking pho tos/videos, repairing houses, and/or preparing for partie s at some spatial locations. Recently, a new framework, namely spatial crowdsourcing[16], for employing workers to conduct spatial tasks, has emerged in both academia (e.g., the datab ase community [8]) and industry (e.g., TaskRabbit [3]). A typic al spatial crowdsourcing platform (e.g., gMission [8] and Med iaQ [17]) assigns a number of moving workersto do spatial tasks nearby, which requires workers to physically move to some specified locations and accomplish these tasks. Note that, not all spatial tasks are as simple as taking a photo or video clip (e.g., street view of Google Maps [2]), monitoring traffic conditions (e.g., Waze [4]), or reportin g local hot spots (e.g., Foursquare [1]), which can be easily completed by providing answers via camera, sensing devices in smart phones, or naked eyes, respectively. In contrast, s ome spatial tasks can be rather complex, such as repairing a hous e, preparing for a party, and performing entertainment shows f r a ceremony, which may consist of several steps/phases/aspec ts, and require demanding professional skills from workers. In other words, these complex tasks cannot be simply accomFig. 1: An Example of Repairing a House in the Multi-Skill Spatial Crowdsourcing System. TABLE I: Worker/Task Skills worker/task skill key set w1, w4 w8 {a1, a4, a6} w2 {a5} w3, w7 {a2, a3} w5, w6 {a1, a5} t1, t2, t3 {a1 ∼ a6} TABLE II: Descriptions of Skills skill key skill description a1 painting walls a2 repairing roofs a3 repairing floors a4 installing pipe systems a5 installing electronic components a6 cleaning plished by normal workers, but require the skilled workers w ith specific expertise (e.g., fixing roofs or setting up the stage ). Inspired by the phenomenon of complex spatial tasks, in this paper, we will consider an important problem in the spatial crowdsourcing system, namely multi-skill spatial crowdsourcing(MS-SC), which assigns multi-skilled workers to those complex tasks, with the matching skill sets and high scores of the worker-and-task assignments. In the sequel, we will illustrate the MS-SC problem by a motivation example of repairing a house. Example 1 (Repairing a House).Consider a scenario of the spatial crowdsourcing in Figure 1, where a user wants to repair a house he/she just bought, in order to have a good living environment for his/her family. However, it is not an easy ta sk to repair the house, which requires many challenging works (skills), such as repairing roofs/floors, replacing/insta lling pipe systems and electronic components, painting walls, and fina lly cleaning rooms. There are many skilled workers that can accomplish one or some of these skill types. In this case, the user can post a spatial task t1, as shown in Figure 1, in the spatial crowdsourcing system, which specifies a set of requi d skills (given in Tables I and II) for the house-repairing tas k, a valid time period to repair, and the maximum budget that he/she would like to pay. In Figure 1, around the spatial location of task t1, there are 8 workers,w1 ∼ w8, each of whom has a different set of skills as given in Table I. For example, worker w1 has the skill set{painting walls, installing pipe systems , cleaning}. To accomplish the spatial task t1 (i.e., repair the house), the spatial crowdsourcing platform needs to select a best subse t of workerswi (1 ≤ i ≤ 8), such that the union of their skill sets can cover the required skill set of task t1, and, moreover, workers can travel to the location of t1 with the maximum net payment under the constraints of arrival times, workers ’ moving ranges, and budgets. For example, we can assign task t1 with 3 workersw2, w7, andw8, who are close tot1, and whose skills can cover all the required skills of t1. Motivated by the example above, in this paper, we will formalize the MS-SC problem, which aims to efficiently assig n workers to complex spatial tasks, under the task constraint s of valid time periods and maximum budgets, such that the required skill sets of tasks are fully covered by those assig ned workers, and the total score of the assignment (defined as the total profit of workers) is maximized. Note that, existing works on spatial crowdsourcing focused on assigning workers to tasks to maximize the total number of completed tasks [16], the number of performed tasks for a worker with an optimal schedule [12], or the reliabilityand-diversity score of assignments [10]. However, they did not take into account multi-skill covering of complex spati al tasks, time/distance constraints, and the assignment scor e with respect to task budgets and workers’ salaries (excluding th e traveling cost). Thus, we cannot directly apply prior solut i ns to solve our MS-SC problem. In this paper, we first prove that our MS-SC problem in the spatial crowdsourcing system is NP-hard, by reducing it from the Set Cover Problem(SCP) [15]. As a result, the MS-SC problem is not tractable, and thus very challenging to achieve the optimal solution. Therefore, in this paper, w e will tackle the MS-SC problem by proposing three effective approximation approaches, greedy, g-divide-and-conquer(gD&C), and cost-model-based adaptive algorithms, which can efficiently compute worker-and-task assignment pairs with the constraints/goals of skills, time, distance, and budgets. Specifically, we make the following contributions. • We formally define themulti-skill spatial crowdsourcing (MS-SC) problem in Section II, under the constraints of multi-skill covering, time, distance, and budget for spati al workers/tasks in the spatial crowdsourcing system. • We prove that the MS-SC problem is NP-hard, and thus intractable in Section II-D. • We propose efficient approximation approaches, namely greedy,g-divide-and-conquer, and cost-model-based adaptive algorithms to tackle the MS-SC problem in Sections IV, V, and VI, respectively. • We conduct extensive experiments on real and synthetic data sets, and show the efficiency and effectiveness of our MSSC approaches in Section VII. Section III introduces a general framework for our MSSC problem in spatial crowdsourcing systems. Section VIII reviews previous works on spatial crowdsourcing. Finally, Section IX concludes this paper. II. PROBLEM DEFINITION In this section, we present the formal definition of the multi-skill spatial crowdsourcing, in which we assign mult iskilled workers with time-constrained complex spatial tas ks. A. Multi-Skilled Workers We first define the multi-skilled workers in spatial crowdsourcing applications. Assume that Ψ = {a1, a2, ..., ak} is a universe ofk abilities/skills. Each worker has one or multiple skills in Ψ, and can provide services for spatial tasks that require some skills inΨ. Definition 1: (Multi-Skilled Workers) Let Wp = {w1, w2, ..., wn} be a set ofn multi-skilled workers at timestamp. Each workerwi (1 ≤ i ≤ n) has a set,Xi (⊆ Ψ), of skills, is located at positionli(p) at timestampp, can move with velocity vi, and has a maximum moving distance di. In Definition 1, the multi-skilled workerswi can move dynamically with speedvi in any direction, and at each timestampp, they are located at spatial places li(p), and prefer to move at most di distance fromli(p). They can freely join or leave the spatial crowdsourcing system. Moreover, each wor ker wi is associated with a set, Xi, of skills, such as taking photos, cooking, and decorating rooms. B. Time-Constrained Complex Spatial Tasks Next, we define complex spatial tasks in the spatial crowdsourcing system, which are constrained by deadlines o f arriving at task locations and budgets. Definition 2: (Time-Constrained Complex Spatial Tasks ) Let Tp = {t1, t2, ..., tm} be a set of time-constrained complex spatial tasks at timestamp. Each tasktj (1 ≤ j ≤ m) is located at a specific location lj, and workers are expected to reach the location of tasktj before the arrival deadlinej. Moreover, to complete the task tj , a set,Yj (⊆ Ψ), of skills is required for those assigned workers. Furthermore, each tas k tj is associated with a budget, Bj , of salaries for workers. As given in Definition 2, usually, a task requester creates a time-constrained spatial task tj , which requires workers physically moving to a specific location lj, and arriving atlj before the arrival deadlinej . Meanwhile, the task requester also specifies a budget, Bj , of salaries, that is, the maximum allowance that h