th
5 International Advanced Technologies Symposium (IATS’09), May 13-15, 2009, Karabuk, Turkey
AGENT-BASED ROUTE PLANNING FOR A MOBILE ROBOT a,*
Uğur Gürela, *, Osman Parlaktunab , Hilal Ezercan Kayır b Eskişehir Osmangazi Üniversitesi Bilgisayar Mühendisliği, Türkiye, E-mail:
[email protected] b Eskişehir Osmangazi Üniversitesi Elektrik Elektronik Mühendisliği, Türkiye E-mail:
[email protected],
[email protected]
Abstract Constructing architecture and forming optimal paths for mobile robots are some of the heavily studied subjects in mobile robot applications. The aim of this paper is to find a sub-optimum path for a single mobile robot using agentbased client-server architecture in a known environment. The sub-optimum path is determined by a heuristic approach, A-Star algorithm. Client-server architecture is formed based on open agent architecture to control the communication between the robot (client) and the server. In the study robots communicate with the server to receive the planned route for the desired starting and the goal point. Simulations on MobileSim simulator program are conducted to show the effectiveness of the proposed architecture.
tasks[4] . Cowlagi and Tsiotras, used agent approach for path planning problem based on Wavelets [5] In this paper, an agent-based approach is proposed for real-time path planning for mobile robot. The approach consists of two modules. The first module is responsible for the path construction, and the second one is responsible for low-level control of the mobile robot. Finding the optimal path solution is time-consuming. For this reason a heuristic approach A-star is used. These two modules communicate via a communication module which is based on open agent architecture (OAA) [6]. The structure of the paper is as follows: In section two proposed method, in section three, application and in section four results and future work is going to be mentioned.
Keywords: Mobile robot, agent, path planning
2. Proposed Method 1. Introduction In autonomous mobile robot domain, applications generally depend on navigating from a known starting point to known goal point in a known or unknown environment. In daily life robots are started to use in many missions like servicing patients in hospitals or delivering postage and messages in offices [1]. In these kinds of applications the points that should be visited are known. But, for a mobile robot, consuming less energy is very important [2] and it is obvious that the more autonomous mobile robot travels the more energy is consumed. In order to consume less energy, a shorter path should be preferred rather than a longer path. Therefore an intelligent path planning algorithm is required. Since the robot works in a real and dynamic environment, the path planning algorithm should construct the path in real time. For this reason a heuristic approach, A-star is used to construct the path. Agent-based approaches also receive considerable attention in the literature. In agent based approaches applications are distributed among several agents. Advantages of this approach are; increase in modularity and increase in robustness of the entire system. Zavlanos and Pappas proposed a method to solve multi-agent assignment problems, in the nature of mobile robots [3]. Rushan et-al applied agent approach to a heterogeneous mobile robot team to find a solution for localization problem. In their work some agents are qualified as localizers. These agents localize themselves and the other mobile robots in the environment. This localization information is passed to all members of the team. The other robots use this information while they perform their
© IATS’09, Karabük University, Karabük, Turkey
In this work basically two kinds of agents are present. Path construction agent is used to construct a path using well known heuristic A-star. The other agent is the mobile robot itself (mobile agent). Path construction agent and mobile agent are communicate with each other through Open Agent Architecture (OAA) framework. (Figure 1)
Figure 1: Communication Framework Open Agent Architecture is a framework for integrating a community of heterogeneous software agents in a distributed environment [7]. The structure of the architecture is shown in Figure 2. In the OAA messaging is made through ICL messages. The ICL messages contain the sender, the receiver and the message information in the message body. All of the messages are flow through the facilitator agent. The Facilitator agent is a pre constructed agent in the OAA environment. ICL message can be constructed as follows: ICLTerm *list=icl_NewList(NULL); ICLTerm *RX=icl_NewInt(rx); icl_AddToList(list,icl_NewInt(rx),TRUE); goal = icl_NewStruct("rota", 1, list);
Gürel, U., Parlaktuna O., Kayır H. E.
Figure 2: OAA structure used in the problem As seen from figure 2, robot agent sends the present coordinates and the goal coordinates to path construction agent through the communication channel and asks a planned path from the path construction agent. The path construction agent plans a path based on the map of the environment. While constructing the path, path planning agent uses A-star algorithm. This algorithm is explained below. A-star search is one of the widely used informed search strategies. It is used to find a path from the starting node to the goal node in a graph. In this study, the environment is divided into 1m. by 1m. square grids. The center point of each square is considered as a node of the graph. Then Astar uses this graph to construct the path. The cost for each grid is calculated as the sum of two costs:
while O is not empty Find nbest from O such that f(nbest) ≤ f(n) for all n in O Remove nbest from O Add nbest to C if nbest = ngoal EXIT end Expand nbest for all x, neighour of nbest and not in C if x is not in O Add x to O elseif g(nbest) + dist(nbest,x)< g(x) Update x’s back-pointer to point the nbest end end The neighborhood of a grid cell uses eight-point connectivity relation and the heuristic distance (h(n) ) is calculated by using Euclidian distance which is always smaller than or equal to the actual distance.
3. Application The explained approach is tested in MobileSim simulation environment. MobileSim is a test-bed for Pioneer robots and the programs are written with Aria framework. The programs that work correctly in simulator also work in the real life. Working environment is drawn by Active Media Mapper program and imported to MobileSim program. [9] The model of the working environment is shown in Figure 3.
f(n) = h(n) + g(n) Here g(n) is the cost to reach from the starting node to the node n , h(n) is the cost to reach from the node n to the goal node. Since g(n) gives the path cost from the start node to node n and h(n) is the heuristic distance which is the estimated cost of the closest path from n to goal, f(n) becomes the estimated cost of the shortest solution through node n. Astar is an optimal search strategy if h(n) is an admissible heuristic that is provided that h(n) never overestimates the exact cost to reach the goal. The input for A-star search algorithm is the graph and the output is a back-pointer path which is a sequence of nodes starting from the goal and back to the start. In the case of O is the open set which is a priority queue and C is the closed set containing all processed nodes, the A-star search algorithm can be expressed as below [8].
Figure 3: Working Environment At the first stage, the mobile agent gives the current grid position and the goal grid position to the path planning agent as shown in Figure 4.
Figure 4: Message of the Robot Agent
Gürel, U., Parlaktuna O., Kayır H. E.
Using this information and the map of the environment, the A-Star algorithm constructs a path between the starting point and the goal point. And the constructed path is passed through communication channel to the robot. The trace of the path followed by the robot is shown in figure 5
4 Conclusions and Future Work In this paper, an agent-based real-time path planning algorithm is proposed for mobile robots. The path is constructed by A-star heuristic. The proposed approach is tested on MobileSim simulation program. Initial results are promising. As shown from the figures, robot navigates from the starting to the goal point. We plan to extend this study for the dynamically changed environments. Also we plan to increase the grid size resolution so that robot can travel in more complicated environments. Another future work would be application of this approach to real world robot applications.
References [1] O. Parlaktuna, A. Sipahioglu, A. Yazici, and U. Gurel, TSP Approach for Mobile Robot Dynamic Path Planning, The First Int. Conf. on Control and Optimization with Industrial Applications(COIA-2005), Baku, Azerbaijan, May 2005. [2] Yongguo M., Yung-Hsiang L ., Y. Charlie Hu, andC.S. George Lee “Energy-Efficient Motion Planning for Mobile Robots”. Proceedings of IEEE International Conference on Robotics and Automation, Volume 5, 4344 – 4349, 26 April-1 May 2004. Figure 5: Trace of the robot motion in simulation 1 Another simulation result with different environment can be seen in Figure 6.
[3] Zavlanos, M.M.; Pappas, G.J.; Robotics, IEEE Transactions on [see also Robotics and Automation, IEEE Transactions on] Volume 24, Issue 1, Feb. 2008 Page(s):232 - 242 Digital Object Identifier 10.1109/TRO.2007.913992 [4] Rushan, S.M.; Mehrandezh, M.; Paranjape, R.; Electrical and Computer Engineering, 2006. CCECE '06. Canadian Conference on May 2006 Page(s):1522 – 1525 Digital Object Identifier 10.1109/CCECE.2006.277605 [5] Cowlagi, R.V.; Tsiotras, P.; American Control Conference, 2008 11-13 June 2008 Page(s):1220 – 1225 Digital Object Identifier 10.1109/ACC.2008.4586659 [6] http://www.ai.sri.com/~oaa/ [7] Xudong Ma; Xianzhong Dai; Dongyao Wang; Xin Jin; Intelligent Robots and Systems, 2005. (IROS 2005). 2005 IEEE/RSJ International Conference on 2-6 Aug. 2005 Page(s):2154 - 2159 Digital Object Identifier 10.1109/IROS.2005.1545101
Figure 6: Trace of the robot motion in simulation 2
[8] Principles of Robot Motion Choset, Howie; Lynch, Kevin M.; Hutchinson, Seth; Kantor, George; Burgard, Wolfram; Kavraki, Lydia E.; Thrun, Sebastian. Al, M.I.T. Press ,2005 pp 527-532 [9] www.activrobots.com/SOFTWARE/aria.html