Kostas E. Bekris

Indoor Human Navigation Systems: A Survey

Whereas outdoor navigation systems typically rely upon GPS, indoor systems have to rely upon dierent techniques for localizing the user, as GPS signals cannot be received indoors. Over the past decade various indoor navigation systems have been developed. This paper provides a comprehensive overview of existing indoor navigation systems and analyzes the dierent techniques used for: (1) locating the user; (2) planning a path; (3) representing the environment; and (4) interacting with the user. Our survey identies a number of research issues that could facilitate large scale deployment of indoor navigation systems.

A Dataset for Improved RGBD-Based Object Detection and Pose Estimation for Warehouse Pick-and-Place

An important logistics application of robotics involves manipulators that pick-and-place objects placed in warehouse shelves. A critical aspect of this task corresponds to detecting the pose of a known object in the shelf using visual data. Solving this problem can be assisted by the use of an RGBD sensor, which also provides depth information beyond visual data. Nevertheless, it remains a challenging problem since multiple issues need to be addressed, such as low illumination inside shelves, clutter, texture-less and reflective objects as well as the limitations of depth sensors. This letter provides a new rich dataset for advancing the state-of-the-art in RGBD-based 3D object pose estimation, which is focused on the challenges that arise when solving warehouse pick-and-place tasks. The publicly available dataset includes thousands of images and corresponding ground truth data for the objects used during the first Amazon Picking Challenge at different poses and clutter conditions. Each image is accompanied with ground truth information to assist in the evaluation of algorithms for object detection. To show the utility of the dataset, a recent algorithm for RGBD-based pose estimation is evaluated in this letter. Given the measured performance of the algorithm on the dataset, this letter shows how it is possible to devise modifications and improvements to increase the accuracy of pose estimation algorithms. This process can be easily applied to a variety of different methodologies for object pose detection and improve performance in the domain of warehouse pick-and-place.

Integrated online localization and navigation for people with visual impairments using smart phones

Indoor localization and navigation systems for individuals with visual impairments (VI) typically rely upon extensive augmentation of the physical space or heavy, expensive sensors; thus, few systems have been adopted. This work describes a system able to guide people with VI through buildings using inexpensive sensors, such as accelerometers, which are available in portable devices like smart phones. The method takes advantage of feedback from the human user, who confirms the presence of landmarks. The system calculates the users location in real time and uses it to provide audio instructions on how to reach the desired destination. Previous work suggested that the accuracy of the approach depended on the type of directions and the availability of an appropriate transition model for the user. A critical parameter for the transition model is the users step length. The current work investigates different schemes for automatically computing the users step length and reducing the dependency of the approach to the definition of an accurate transition model. Furthermore, the direction provision method is able to use the localization estimate and adapt to failed executions of paths by the users. Experiments are presented that evaluate the accuracy of the overall integrated system, which is executed online on a smart phone. Both people with visual impairments, as well as blindfolded sighted people, participated in the experiments. The experiments included paths along multiple floors, that required the use of stairs and elevators.

Efficient sampling-based motion planning with asymptotic near-optimality guarantees for systems with dynamics

Recent motion planners, such as RRT*, that achieve asymptotic optimality require a local planner, which connects two states with a trajectory. For systems with dynamics, the local planner corresponds to a two-point boundary value problem (BVP) solver, which is not always available. Furthermore, asymptotically optimal solutions tend to increase computational costs relative to alternatives, such as RRT, that focus on feasibility. This paper describes a sampling-based solution with the following desirable properties: a) it does not require a BVP solver but only uses a forward propagation model, b) it employs a single propagation per iteration similar to RRT, making it very efficient, c) it is asymptotically near-optimal, and d) provides a sparse data structure for answering path queries, which further improves computational performance. Simulations on prototypical dynamical systems show the method is able to improve the quality of feasible solutions over time and that it is computationally efficient.

Sparse Methods for Efficient Asymptotically Optimal Kinodynamic Planning

This work describes STABLE SPARSE RRT (SST), an algorithm that (a) provably provides asymptotic (near-)optimality for kinodynamic planning without access to a steering function, (b) maintains only a sparse set of samples, (c) converges fast to high-quality paths and (d) achieves competitive running time to RRT, which provides only probabilistic completeness. SST addresses the limitation of RRT (^*), which requires a steering function for asymptotic optimality . This issue has motivated recent variations of RRT (^*), which either work for a limiting set of systems or exhibit increased computational cost. This paper provides formal arguments for the properties of the proposed algorithm. To the best of the authors’ knowledge, this is the first sparse data structure that provides such desirable guarantees for a wide set of systems under a reasonable set of assumptions. Simulations for a variety of benchmarks, including physically simulated ones, confirm the argued properties of the approach.

PRACSYS: an extensible architecture for composing motion controllers and planners

This paper describes a software infrastructure for developing controllers and planners for robotic systems, referred here as PRACSYS . At the core of the software is the abstraction of a dynamical system, which, given a control, propagates its state forward in time. The platform simplifies the development of new controllers and planners and provides an extensible framework that allows complex interactions between one or many controllers, as well as motion planners. For instance, it is possible to compose many control layers over a physical system, to define multi-agent controllers that operate over many systems, to easily switch between different underlying controllers, and plan over controllers to achieve feedback-based planning. Such capabilities are especially useful for the control of hybrid and cyber-physical systems, which are important in many applications. The software is complementary and builds on top of many existing open-source contributions. It allows the use of different libraries as plugins for various modules, such as collision checking, physics-based simulation, visualization, and planning. This paper describes the overall architecture, explains important features and provides use-cases that evaluate aspects of the infrastructure.

Cloud Automation: Precomputing Roadmaps for Flexible Manipulation

The goal of this article is to highlight the benefits of cloud automation for industrial adopters and some of the research challenges that must be addressed in this process. The focus is on the use of cloud computing for efficiently planning the motion of new robot manipulators designed for flexible manufacturing floors. In particular, different ways that a robot can interact with a computing cloud are considered, where an architecture that splits computation between the remote cloud and the robot appears advantageous. Given this synergistic robot-cloud architecture, this article describes how solutions from the recent literature can be employed on the cloud during a periodically updated preprocessing phase to efficiently answer manipulation queries on the robot given changes in the workspace. In this setup, interesting tradeoffs arise between path quality and computational efficiency, which are evaluated through simulation. These tradeoffs motivate further research on how motion planning should be executed given access to a computing cloud.

Dealing with Difficult Instances of Object Rearrangement

Rearranging multiple objects is a critical skill for robots so that they can effectively deal with clutter in human spaces. This is a challenging problem as it involves combinatorially large, continuous C-spaces involving multiple movable bodies and complex kinematic constraints. This work initially revisits an existing search-based approach, which solves monotone challenges, i.e., when objects need to be grasped only once so as to be rearranged. The first contribution is the extension of this technique to a method that addresses many non-monotone challenges. The second contribution is the use of either the monotone or of the new non-monotone method as a local planner in the context of a higher-level task planner that searches the space of object placements and which provides stronger guarantees. The paper aims to emphasize the benefit of using more powerful motion primitives in the context of task planning for object rearrangement than an individual pick-andplace. Experiments in simulation using a model of a Baxter robot arm show the capability of solving difficult instances of rearrangement problems and evaluate the methods in terms of success ratio, running time, scalability and path quality.

Geometric probability results for bounding path quality in sampling-based roadmaps after finite computation

Sampling-based algorithms provide efficient solutions to high-dimensional, geometrically complex motion planning problems. For these methods asymptotic results are known in terms of completeness and optimality. Previous work by the authors argued that such methods also provide probabilistic near-optimality after finite computation time using indications from Monte Carlo experiments. This work formalizes these guarantees and provides a bound on the probability of finding a near-optimal solution with PRM* after a finite number of iterations. This bound is proven for general-dimension Euclidean spaces and evaluated through simulation. These results are leveraged to create automated stopping criteria for PRM* and sparser near-optimal roadmaps, which have reduced running time and storage requirements.

Planning representations and algorithms for prehensile multi-arm manipulation

This paper describes the topology of general multi-arm prehensile manipulation. Reasonable assumptions are applied to reduce the number of manipulation modes, which results in an explicit graphical representation for multi-arm manipulation that is computationally manageable to store and search for solution paths. In this context, it is also possible to take advantage of preprocessing steps to significantly speed up online query resolution. The approach is evaluated in simulation for multiple arms showing it is possible to quickly compute multi-arm manipulation paths of high-quality on the fly.