No abstract

Map-based navigation is a crucial task for any mobile robot. Usually, in an unknown environment this problem is addressed by applying Simultaneous Localization and Mapping (SLAM) based on metric grid-maps. However, such maps are in general rather computational expensive and do not scale well. Insects are able to cover large distances and reliably find back to their nests, although they are quite limited in their resources. Inspired by theories on insect navigation, we developed a data structure which is highly scalable and efficiently adapts to the available memory during run-time. Positions in space are memorized as snapshots, which are unique configurations of landmarks. Unlike conventional snapshot or visual map approaches, we do not simply store the landmarks as a set, but we arrange them in a tree-like structure according to the relevance of their information. The resulting navigation solely relies on the direction measurements of arbitrary landmarks. In this work we present the concept of the Landmark-Tree map and apply it to a mobile platform equipped with an omnidirectional camera. We verify the reliability and robustness of the LT-map concept in simulations as well as by experiments with the robotic platform.Keywords: topological navigation; roadmap; landmark-tree; LT-map; bio-inspired MotivationAutonomous navigation is a highly complex task, which often requires most resources on mobile robots. The navigation problem can in general be divided into local navigation, where the robot moves within its close surroundings relative to a local frame and solves a specific task, and global navigation, where the robot travels between task-related workspaces, e.g. from its base to the location of interest.Typical local navigation problems include obstacle avoidance, collecting probes and assembling components. To accomplish the respective task, the robot needs to have detailed geometrical knowledge of its workspace for precise localization and for planning accurate trajectories. If no a priori maps are available, the robot has to generate a complete environment model itself.Since robotic systems tend to get smaller and more agile and the number of applications in which robots need to cover long distances is growing, the global navigation problem gains importance. To name a few examples, think of a Micro Aerial Vehicle (MAV), which starts at a rescue team and flies in a specific direction to search for people who require help, or a rover on a foreign planet which has to find back to the base station for analyzing the collected probes. These robots must be able to reliably reach a previously visited, possibly far distant location.

Abstract-Metric maps provide a reliable basis for mobile robot navigation. However, such maps are in general quite resource expensive and do not scale very well. Aiming for a highly scalable map, we adopt theories of insect navigation to develop an algorithm which builds a topological map for global navigation. Similar to insect conduct, positions in space are memorized as snapshots, which are unique configurations of landmarks. Unlike conventional snapshot approaches, we do not simply store the landmarks as a set, but we build a landmark tree which enables us to easily free memory in case of a continuously growing map while still preserving the dominant information. The resulting navigation is not sensor specific and solely relies on the directions of arbitrary landmarks. The generated map enables a mobile robot to navigate between defined locations and let it retrace a previously pursued path. Finally, we verify the reliability of the Landmark-Tree Map (LT-Map) concept and its robustness on memory limitations.

Sequence optimization is an important problem in many production automation scenarios involving industrial robots. Mostly, this is done by reducing it to Traveling Salesman Problem (TSP). However, in many industrial scenarios optimization potential is not only hidden in optimizing a sequence of operations but also in optimizing the individual operations themselves. From a formal point of view, this leads to the Traveling Salesman Problem with Neighborhoods (TSPN). TSPN is a generalization of TSP where areas should be visited instead of points. In this paper we propose a new method for solving TSPN efficiently. We compare the new method to the related approaches using existing test benchmarks from the literature. According to the evaluation on instances with known optimal values, our method is able to obtain a solution close to the optimum.