Martin Krzykawski

Dense topological maps and partial pose estimation for visual control of an autonomous cleaning robot

We present a mostly vision-based controller for mapping and completely covering a rectangular area by meandering cleaning lanes. The robot is guided along a parallel course by controlling the current distance to its previous lane. In order to frequently compute and-if necessary-correct the robots distance to the previous lane, a dense topological map of the robots workspace is built. The map stores snapshots, i.e. panoramic images, taken at regular distances while moving along a cleaning lane. For estimating the distance, we combine bearing information obtained by local visual homing with distance information derived from the robots odometry. In contrast to traditional mapping applications, we do not compute the robots full pose w.r.t. an external reference frame. We rather rely on partial pose estimation and only compute the sufficient and necessary information to solve the task. For our specific application this includes estimates of (i) the robots distance to the previous lane and of (ii) the robots orientation w.r.t. world coordinates. The results show that the proposed method achieves good results with only a small portion of overlap or gaps between the lanes. The dense topological representation of space and the proposed controller will be used as building blocks for more complex cleaning strategies making the robot capable of covering complex-shaped workspaces such as rooms or apartments.

Cleaning robot navigation using panoramic views and particle clouds as landmarks

The paper describes a visual method for the navigation of autonomous floor-cleaning robots. The method constructs a topological map with metrical information where place nodes are characterized by panoramic images and by particle clouds representing position estimates. Current image and position estimate of the robot are interrelated to landmark images and position estimates stored in the map nodes through a holistic visual homing method which provides bearing and orientation estimates. Based on these estimates, a position estimate of the robot is updated by a particle filter. The robots position estimates are used to guide the robot along parallel, meandering lanes and are also assigned to newly created map nodes which later serve as landmarks. Computer simulations and robot experiments confirm that the robot position estimate obtained by this method is sufficiently accurate to keep the robot on parallel lanes, even in the presence of large random and systematic odometry errors. This ensures an efficient cleaning behavior with almost complete coverage of a rectangular area and only small repeated coverage. Furthermore, the topological-metrical map can be used to completely cover rooms or apartments by multiple meander parts.

A Vision-Based Trajectory Controller for Autonomous Cleaning Robots

Autonomous cleaning robots should completely cover the accessible area with minimal repeated coverage. We present a mostly visionbased navigation strategy for systematical exploration of an area with meandering lanes. The results of the robot experiments show that our approach can guide the robot along parallel lanes while achieving a good coverage with only a small proportion of repeated coverage. The proposed method can be used as a building block for more elaborated navigation strategies which allow the robot to systematically clean rooms with a complex workspace shape.

Parsimonious loop-closure detection based on global image-descriptors of panoramic images

In the context of vision-based topological navigation, detecting loop closures requires to compare the robots current camera image to a large number of images stored in the map. For efficient image comparisons, we apply distance functions to global image-descriptors, i.e. low-dimensional descriptors derived from the entire panoramic images. To identify promising combinations of descriptors and distance functions, we formulate the loop-closure detection as a binary classification problem and analyze the resulting receiver operator characteristics (ROC). The results of comparing a wide range of descriptors and distance functions reveal that reliable loop-closure detection is possible with a single 16- to 128-dimensional image-descriptor based on gray-value histograms or Fourier descriptors and that all considered distance functions have a comparable performance.