Stigmergy at Work: Planning and navigation for a service robot on an RFID floor Ali Abdul Khaliq Alessandro Saffiotti AASS Cognitive Robotic Systems Lab, ¨ Orebro University, ¨ Orebro, Sweden Contact email: ali-abdul.khaliq@aass.oru.se Abstract— Many species in nature store information in the environment to facilitate the performance of tasks and enable cooperation. This principle is known as stigmergy. Stigmergy has been widely studied in robotic systems, but so far mostly in simulation or in laboratory proofs of concept. In this paper, we propose a stigmergic approach to goal-directed navigation that can be used for navigation of a full-scale robotic system in a real apartment. A team of small ePuck robots build a set of navigation maps directly onto an RFID floor, where each map is associated to one predefined goal. The information stored in the floor can then used by a mid-size robot or by a larger domestic robot to perform safe navigation toward the predefined goals. To navigate, robots only rely on the information read from the RFID tags: in particular, they do not need to use an internal map or to perform self-localization. This results in robust and repeatable navigation with minimal hardware and software requirements. I. I NTRODUCTION The majority of autonomous mobile robots rely on maps to perform goal-directed navigation. A map is a representation of the working environment, which is stored inside the robot and is used to plan suitable paths to goal locations. Although widely accepted, the use of internal maps entails a number of serious difficulties that make robot navigation still challenging. In order to make use of the map, the robot must relate it to the physical environment: in particular, it must assess the correspondence between representations in the map and features in the environment, and estimate its own position with respect to the map and to the planned path. Self-localization is one of the most studied problems in mobile robotics, and despite the impressive progress it remains the Achilles heel of mobile robotic systems — the component which is most often responsible for failures. In addition, self localization typically relies on expensive long- range sensors, on complex computation, or on both. The use of internal maps also implies that whenever the robot enters a new workspace it must given (or build) a map, together with its own initial position. Finally, if multiple robots operate in the same environment, they need the ability to share their internal maps and to keep them registered. In this paper, we explore an alternative route: to build and maintain the map into the environment itself. We enrich the environment with a grid of read-write RFID tags embedded in the floor, and we provide algorithms by which a robot (or set of robots) can build navigation maps directly into these tags and use these maps for goal-directed navigation. These algorithms have the following features: (1) the only required sensors on the robot are an RFID tag reader and an obstacle sensor; (2) in particular, we do not assume that the robot is able to assess its position or orientation; (3) the algorithms use minimal computational resources; (4) they can be run cooperatively by multiple robots; (5) the maps are stored permanently in the environment and can be used by any robot; and (6) the resulting navigation paths are optimal. The idea to store the map in the environment is not new. Brooks [1] famously argued that “the world is its own best model”; in a similar vein, in our approach we augment the world by storing a model into it. From a computational perspective, our approach relies on the concepts of external- ization and of spatialization. Externalization means that part of the state of computation is stored outside the computing agent, in a medium that can be accessed by this or other agents possibly at a later time. Spatialization means that the information is stored in the environment at the location which it refers to. These concepts are at the basis of the field of spatial computing [2], [3]. In nature, the use of external and spatialized information in the performance of tasks is known as stigmergy. Stigmergy allows insects of same species to communicate indirectly with each other to organize their individual or collective work through traces left in the environment [4]. These traces are typically made of volatile chemicals called pheromones [5]. The environment and pheromones can be seen as an external memory, whose contents are written by an agent and then read by a different agent, or by same agent at a later time, to decide its behavior. Stigmergy has been studied in robotic systems since the early 1990’s [6], [7], [8], [9], and several approaches have been proposed that use RFID tags as artificial pheromone. Some of these [10], [11] store pheromone-based potential fields in RFID tags and use a kernel method to create a navigation gradient. Others [12] use RFID tags to allow humans and robots to mark objects and places, and to leave trails that lead to them. In [13], rescue robots deploy RFID tags during exploration, and later use them for localization. Johansson and Saffiotti [14] define a stigmergic wavefront algorithm to build a goal-directed navigation map on a RFID floor. Successive works [15], [16] extend the latter approach. While the above tradition in stigmergic robotics is rich in ideas, most approaches have only been tested in simulation or in laboratory proofs of concept using miniature robots and small artificial environments. This paper makes one fun- damental step forward: it shows that stigmergic approaches