Dynamic Complex Task Allocation in Multisensor Surveillance Systems Ahmed M. Elmogy, Alaa M. Khamis, and Fakhri O. Karray University of Waterloo, ON, Canada Abstract— In this paper, a centralized and hierarchical dynamic and fixed tree task allocation approaches are presented to solve complex task allocation problem in multisensor surveillance systems. Complex tasks are tasks that can be decomposed to a set of subtasks and so can be executed by several possible ways. The complex task allocation problem is not only concerned with how to assign a set of tasks to a set of mobile robots, but also with how to coordinate the behavior of these mobile sensing nodes in order to perform cooperative tasks efficiently. The simulation results show that hierarchical dynamic tree task allocation outperforms all the other techniques especially in complex surveillance operations where large number of robots is used to scan large number of areas. Index Terms—Mobile sensors, Surveillance Systems, Task Allocation, Market-based Approach. I. INTRODUCTION A UTONOMOUS multi-robot systems have become an active research area and are highly seen in several new application areas in the recent years. These applications include but are not limited to space and underwater exploration, search and rescue, humanitarian mining, surveillance and reconnaissance, etc. Many benefits can be anticipated from the use of multi-robot system such as resolving complexity by decomposing the complex task into simple tasks, decreasing task completion time, and increasing mission reliability. Thus having only one robot may work as a bottleneck for the whole system especially in critical times. One of the challenging domain for multi-robot systems is surveillance. Advanced surveillance systems include a vast array of cooperative (static and mobile) sensors with varying sensing modalities that can sense continuously the volume of interest [1]. In these surveillance systems, mobile sensors are basically a multi-robot system that is used as a complementary solution to overcome the limitations of static sensors. Ideally robots in the multi-robot system will coordinate to distribute the tasks amongst themselves in a way that enables them to accomplish their mission efficiently and reliably [2,3]. This makes dynamic task allocation one of the essential requirements for mobile surveillance systems. The past decade has witnessed a growing emphasis in research topics highlighting multi-robot task allocation (MRTA) [3-6]. In spite of the great number of MRTA algorithms reported in the literature, important aspects have, to date been given little attention. These aspects include but are not restricted to allocation of complex tasks, dynamic task allocation, and constrained task allocation. In this paper, we are trying to address these aspects by giving an organizational framework to study this problem in a formal manner. Market-based approaches have received significant attention and are growing very fast in the last few decades especially in multi-agent domains [7,8,9,10]. These approaches are considered as hybrid approaches that combine the centralized and hierarchical strategies. Motivated by this great attention, a hierarchical market-based architecture for complex task allocation is presented in this paper. Complex tasks are tasks that can be decomposed into subtasks. The proposed architecture integrates low-level motion control with high- level task allocation for mobile sensor network. In order to reach the low-level motion control design, a traveling salesman path planning technique is used. Fixed and dynamic task trees are used as implementations of tasks, which are allocated to robots using auctioning. Different types of auctions are introduced. The rest of this paper is organized as follows. Section 2 introduces the task allocation problem followed by summarizing the related work reported in the literature to solve this problem in sections 3. Section 4 describes the proposed algorithms for task allocation in details. Section 5 reports our simulations on fixed and dynamic tree task allocations. A comparison with other algorithms is also introduced in this section. Finally, conclusion and future work are summarized in section 6. II. PROBLEM DEFINITION Generally, the problem of task allocation is to find the optimal allocation of a set of tasks T to a subset of robots R , which will be responsible for accomplishing it [9] : A T R  (1) Where A is the allocation function, which maps each task or set of tasks T to one robot or subsets of robots R . The goal is to assign robots to tasks so as to maximize the overall performance of the system, which will be governed by the utility function: 1 1 n m U ij i j      u (2) Where n, and m are the number of mobile sensors and tasks respectively. is the utility of executing task j by a mobile robot i . u ij Each mobile sensor r R  can express its ability to execute a task t T  , or a bundle of tasks B T  through bids () r b t or ( ) r b B . The cost of a bundle of tasks can be simply: ( ) ( ) { } 1 f b B b t t B r r s s s     (3) Where f is the number of tasks of the bundle B . The group’s assignment determines the bundle B T  of tasks that each mobile sensor r R  receives. The global objective function can vary depending on the requirements of the system or the preferences of the designer. The most common global 2009 International Conference on Signals, Circuits and Systems -1- 978-1-4244-4398-7/09/$25.00 ©2009 IEEE