Skip to main content

Über dieses Buch

This volume presents a collection of papers presented at the 16th International Symposium of Robotic Research (ISRR). ISRR is the biennial meeting of the International Foundation of Robotic Research (IFRR) and its 16th edition took place in Singapore over the period 16th to 19th December 2013. The ISRR is the longest running series of robotics research meetings and dates back to the very earliest days of robotics as a research discipline. This 16th ISRR meeting was held in the 30th anniversary year of the very first meeting which took place in Bretton Woods (New Hampshire, USA) in August 1983., and represents thirty years at the forefront of ideas in robotics research.

As for the previous symposia, ISRR 2013 followed up on the successful concept of a mixture of invited contributions and open submissions. 16 of the contributions were invited contributions from outstanding researchers selected by the IFRR officers and the program committee, and the other contributions were chosen among the open submissions after peer review. This selection process resulted in a truly excellent technical program which featured some of the very best of robotic research. These papers were presented in a single-track interactive format which enables real conversations between speakers and the audience.

The symposium contributions contained in this volume report on a variety of new robotics research results covering a broad spectrum organized into traditional ISRR categories: control; design; intelligence and learning; manipulation; perception; and planning.





Is Active Impedance the Key to a Breakthrough for Legged Robots?

This work addresses the question whether active impedance control is key to a breakthrough for legged robots. In this paper, we will talk about controlling the mechanical impedance of joints and legs with a focus on stiffness and damping control. In contrast to passive elements like springs, active impedance is achieved by torque-controlled joints allowing real-time adjustment of stiffness and damping. We argue that legged robots require a high degree of versatility and flexibility to execute a wide range of assistive tasks to be truly useful to humans and thus to lead to a breakthrough. Adjustable stiffness and damping in realtime is a fundamental building block towards versatility. Experiments with our 80 kg hydraulic quadruped robot HyQ demonstrate that active impedance alone (thus no springs in the structure) can successfully emulate passively compliant elements during highly-dynamic locomotion tasks (running and hopping); and, that no springs are needed to protect the actuation system. Here we present results of a flying trot, also referred to as running trot. To the authors’ best knowledge this is the first time a flying trot was successfully implemented on a robot without passive elements such as springs. A critical discussion on the pros and cons of active impedance concludes the paper. An extended version of this paper has been published in IJRR in 2015 [43].

Claudio Semini, Victor Barasuol, Thiago Boaventura, Marco Frigerio, Jonas Buchli

Optimal Control of Nonlinear Systems with Temporal Logic Specifications

We present a mathematical programming-based method for optimal control of nonlinear systems subject to temporal logic task specifications. We specify tasks using a fragment of linear temporal logic (LTL) that allows both finite- and infinite-horizon properties to be specified, including tasks such as surveillance, periodic motion, repeated assembly, and environmental monitoring. Our method directly encodes an LTL formula as mixed-integer linear constraints on the system variables, avoiding the computationally expensive process of creating a finite abstraction. Our approach is efficient; for common tasks our formulation uses significantly fewer binary variables than related approaches and gives the tightest possible convex relaxation. We apply our method on piecewise affine systems and certain classes of differentially flat systems. In numerical experiments, we solve temporal logic motion planning tasks for high-dimensional (10$$+$$+ continuous state) systems.

Eric M. Wolff, Richard M. Murray

Extended LQR: Locally-Optimal Feedback Control for Systems with Non-Linear Dynamics and Non-Quadratic Cost

We present Extended LQR, a novel approach for locally-optimal control for robots with non-linear dynamics and non-quadratic cost functions. Our formulation is conceptually different from existing approaches, and is based on the novel concept of LQR-smoothing, which is an LQR-analogue of Kalman smoothing. Our approach iteratively performs both a backward Extended LQR pass, which computes approximate cost-to-go functions, and a forward Extended LQR pass, which computes approximate cost-to-come functions. The states at which the sum of these functions is minimal provide an approximately optimal sequence of states for the control problem, and we use these points to linearize the dynamics and quadratize the cost functions in the subsequent iteration. Our results indicate that Extended LQR converges quickly and reliably to a locally-optimal solution of the non-linear, non-quadratic optimal control problem. In addition, we show that our approach is easily extended to include temporal optimization, in which the duration of a trajectory is optimized as part of the control problem. We demonstrate the potential of our approach on two illustrative non-linear control problems involving simulated and physical differential-drive robots and simulated quadrotor helicopters.

Jur van den Berg

Adaptive Communication in Multi-robot Systems Using Directionality of Signal Strength

We consider the problem of satisfying communication demands in a multi-agent system where several robots cooperate on a task and a fixed subset of the agents act as mobile routers. Our goal is to position the team of robotic routers to provide communication coverage to the remaining client robots. We allow for dynamic environments and variable client demands, thus necessitating an adaptive solution. We present an innovative method that calculates a mapping between a robot’s current position and the signal strength that it receives along each spatial direction, for its wireless links to every other robot. We show that this information can be used to design a simple positional controller that retains a quadratic structure, while capturing the behavior of wireless signals in real-world environments. Notably, our approach does not necessitate stochastic sampling along directions that are counter-productive to the overall coordination goal, nor does it require exact client positions, or a known map of the environment.

Stephanie Gil, Swarun Kumar, Dina Katabi, Daniela Rus

Multi-vehicle Dynamic Pursuit Using Underwater Acoustics

Marine robots communicating wirelessly is an increasingly attractive means for observing and monitoring the ocean, but acoustic communication remains a major impediment to real-time control. In this paper we address through experiments the capability of acoustics to sustain highly dynamic, multi-agent missions, in particular range-only pursuit in a challenging shallow-water environment. We present in detail results comparing the tracking performance of three different communication configurations, at operating speeds near 1.5 m/s. A “lower bound” case with RF wireless communication, a 4-second cycle and no quantization has a tracking bandwidth of $$\approx $$≈0.5 rad/s. When using full-sized modem packets with negligible quantization and a 23-second cycle time, the tracking bandwidth is $$\approx $$≈0.065 rad/s. With 13-bit mini-packets, we employ logarithmic quantization to achieve a cycle time of 12 s and a tracking bandwidth of $$\approx $$≈0.13 rad/s. These outcomes show definitively that aggressive dynamic control of multi-agent systems underwater is tractable today.

Brooks Reed, Josh Leighton, Milica Stojanovic, Franz Hover

Aggressive Maneuver Regulation of a Quadrotor UAV

In this paper we design a nonlinear controller for aggressive maneuvering of a quadrotor. We take a maneuver regulation perspective. Differently from the classical trajectory tracking approach, maneuver regulation does not require following a timed reference state, but a geometric “path” with a velocity (and possibly orientation) profile assigned on it. The proposed controller relies on three main ideas. Given a desired maneuver, i.e., a set of state trajectories equivalent under time translations, the system dynamics is decomposed into dynamics longitudinal and transverse to the maneuver. A space-dependent version of the transverse dynamics is derived, by using the longitudinal state, i.e., the arc-length of the path, as an independent variable. Then the controller is obtained as a function of the arc-length consisting of two terms: a feedforward term, being the nominal input to apply when on the path at the current arc-length, and a feedback term exponentially stabilizing the state-dependent transverse dynamics. Numerical computations are presented to prove the effectiveness of the proposed strategy. The controller performances are tested in presence of uncertainty of the model parameters and input noise and saturations. The controller is also tested in a realistic simulation environment validated against an experimental test-bed.

Sara Spedicato, Giuseppe Notarstefano, Heinrich H. Bülthoff, Antonio Franchi

Towards Modeling Real-Time Trust in Asymmetric Human–Robot Collaborations

We are interested in enhancing the efficiency of human–robot collaborations, especially in “supervisor-worker” settings where autonomous robots work under the supervision of a human operator. We believe that trust serves a critical role in modeling the interactions within these teams, and also in streamlining their efficiency. We propose an operational formulation of human–robot trust on a short interaction time scale, which is tailored to a practical tele-robotics setting. We also report on a controlled user study that collected interaction data from participants collaborating with an autonomous robot to perform visual navigation tasks. Our analyses quantify key correlations between real-time human–robot trust assessments and diverse factors, including properties of failure events reflecting causal trust attribution, as well as strong influences from each user’s personality. We further construct and optimize a predictive model of users’ trust responses to discrete events, which provides both insights on this fundamental aspect of real-time human–machine interaction, and also has pragmatic significance for designing trust-aware robot agents.

Anqi Xu, Gregory Dudek

Optimal Control for Viscoelastic Robots and Its Generalization in Real-Time

Inspired by the elasticity contained in human muscles and tendons, viscoelastic joints are designed with the aim of imitating human motions by exploiting their ability to mechanically store and release potential energy. This distinct feature makes elastic robots especially interesting to the application of optimal control principles, as generating such motions is not possible by data-driven paradigms. In particular, reaching peak velocities by using the stored energy in the springs is of great interest, as such capabilities might open up entirely new application domains. In this paper, we review our results on solving various optimal control problems for elastic joints and full scale robot arms, as well as the experimental validation. Clearly, solving optimal control problems for highly nonlinear full robot dynamics is feasible nowadays only numerically, i.e. offline. In turn, optimal solutions would only contribute a clear benefit for real tasks, if they would be accessible/generalizable in real-time. For this, we developed a framework for executing near-optimal motions of elastic robot arms in real-time. In contrast to existing approaches, we use dynamically optimal motions (i.e. offline solutions of optimal control problems) as given learning input and then apply generalization via Dynamic Movement Primitives (DMPs). With this approach, we intend to overcome the well-known problems of optimal control and data-driven learning with associated generalization: being offline and being suboptimal (In fact, data-driven approaches can only be applied if the solution is already quite obvious for the human teacher. In case of highly nonlinear problems these “intuitive” initial solutions are typically not available.), respectively.

Sami Haddadin, Roman Weitschat, Felix Huber, Mehmet Can Özparpucu, Nico Mansfeld, Alin Albu-Schäffer

K-Redundant Trees for Safe and Efficient Multi-robot Recovery in Complex Environments

This paper presents a self-stabilizing distributed algorithm to recover a large number of robots safely and efficiently in a goal location. Previously, we designed a distributed algorithm, called DMLST, to recover robots [1]. Our approach constructed a maximum-leaf spanning tree for physical routing, such that interior robots remained stationary and leaf robots move. In this paper, we extend our approach to k-DMLST recovery that provides k-connectivity in the network, meaning that each robot is connected to the goal location through k trees. This redundancy provides stronger network connectivity by reducing the probability of losing the parent during recovery. We also propose an efficient navigation algorithm for the motion of robots which guarantees forward progress during the recovery. k-DMLST recovery has been tested and compared with other methods in simulation, and implemented on a physical multi-robot system. A basic recovery algorithm fails in all experiments, and DMLST recovery is not successful in few trials. However, k-DMLST recovery efficiently recovers more than 90 % of robots in all trials.

Golnaz Habibi, Lauren Schmidt, Mathew Jellins, James McLurkin

Adaptive Inter-Robot Trust for Robust Multi-Robot Sensor Coverage

This paper proposes a new approach to both characterize inter-robot trust in multi-robot systems and adapt trust online in response to the relative performance of the robots. The approach is applied to a multi-robot coverage control scenario, in which a team of robots must spread out over an environment to provide sensing coverage. A decentralized algorithm is designed to control the positions of the robots, while simultaneously adapting their trust weightings. Robots with higher quality sensors take charge of a larger region in the environment, while robots with lower quality sensors have their regions reduced. Using a Lyapunov-type proof, it is proven that the robots converge to locally optimal positions for sensing that are as good as if the robots’ sensor qualities were known beforehand. The algorithm is demonstrated in Matlab simulations.

Alyssa Pierson, Mac Schwager



AIST Humanoid Robotics Challenge

This paper presents the challenges assigned to humanoid robots by National Institute of Advanced Industrial Science and Technology (AIST). Since the Ministry of Economy, Trade and Industry of Japan initiated a research and development (R&D) project on humanoid robotics, AIST has developed platforms for humanoid robotics research and conducted R&D on humanoid robots applications. Particular attention has been developed to the requirements of industries using the platforms. Following TEPCO’s Fukushima Daiichi Nuclear Power Station accident in 2011, DARPA started the DARPA Robotics Challenge (DRC) to promote innovation in robotic technology for disaster-response operations. In the DRC, participating robots are assigned a series of challenging tasks that indicate their suitability for disaster response. This paper presents AIST Humanoid Challenge related the DRC-selected tasks.

Kazuhito Yokoi

A Scripted Printable Quadrotor: Rapid Design and Fabrication of a Folded MAV

Robotic systems hold great promise to assist with household, educational, and research tasks, but the difficulties of designing and building such robots often are an inhibitive barrier preventing their development. This paper presents a framework in which simple robots can be easily designed and then rapidly fabricated and tested, paving the way for greater proliferation of robot designs. The Python package presented in this work allows for the scripted generation of mechanical elements, using the principles of hierarchical structure and modular reuse to simplify the design process. These structures are then manufactured using an origami-inspired method in which precision cut sheets of plastic film are folded to achieve desired geometries. Using these processes, lightweight, low cost, rapidly built quadrotors were designed and fabricated. Flight tests compared the resulting robots against similar micro air vehicles (MAVs) generated using other processes. Despite lower tolerance and precision, robots generated using the process presented in this work took significantly less time and cost to design and build, and yielded lighter, lower power MAVs.

Ankur M. Mehta, Daniela Rus, Kartik Mohta, Yash Mulgaonkar, Matthew Piccoli, Vijay Kumar

The Solving by Building Approach Based on Thermoplastic Adhesives

While, in nature, changes of morphology such as body shape, size, and strength play essential roles in animals’ adaptability in a variety of environment, our robotic systems today still severely suffer from the lack of flexibility in morphology which is one of the most significant bottlenecks for their autonomy and adaptability. With the ability to autonomously modify own body shapes or mechanical structures in surroundings, robotic systems could achieve a variety of tasks in flexible and simple manners. For this reason, we have been investigating technological solutions based on a class of unconventional material, the so-called Thermoplastic Adhesives (TPAs), with which the robots are able to construct their own body parts as well as connecting and disconnecting various mechanical structures. Based on our technological exploration so far, in this paper, we introduce the concept of “solving-by-building” approach, in which we consider how autonomous construction of mechanical parts can help robots to improve performances or to “solve” problems in given tasks. Unlike the conventional adaptive systems that can only learn motor control policies, the ability to change mechanical structures can potentially deal with a significantly more variety of problems. By introducing some of the recent case studies in our laboratory, we discuss the challenges and perspectives of the solving-by-building approach based on TPAs.

Fumiya Iida, Liyu Wang, Luzius Brodbeck, Derek Leach, Surya Nurzaman, Utku Culha

Slip Detection in a Novel Tactile Force Sensor

Tactile sensing improves the manipulation and grasping of unknown objects. It contributes to increase the knowledge of the environment and provides useful information to improve grasping control. The sensors traditionally used for tactile sensing emphasize in grasping object shape and force detection. However slip detection is also crucial to successfully manipulate an object. Several approaches have appeared to detect slipping, the majority being a combination of complex sensors with complex algorithms. In this paper, we present a simple, low cost and durable tactile force sensor and its use to slip detection via a simple but effective method based on micro-vibration detection. We also analyze the results of using the same principle to detect slip in other force sensors based on flexible parts. In particular, we also show the slip detection with: a flexible finger (designed by the authors) acting as a force sensor, the finger torque sensor of a commercial robotic hand (Barrett Hand), and a commercial 6-axis force sensor mounted in the wrist of a robot.

Raul Fernandez, Ismael Payo, Andres S. Vazquez, Jonathan Becedas

Concentric Tube Robots: The State of the Art and Future Directions

Seven years ago, concentric tube robots were essentially unknown in robotics, yet today one would be hard pressed to find a major medical robotics forum that does not include several presentations on them. Indeed, we now stand at a noteworthy moment in the history of these robots. The recent maturation of foundational models has created new opportunities for research in control, sensing, planning, design, and applications, which are attracting an increasing number of robotics researchers with diverse interests. The purpose of this review is to facilitate the continued growth of the subfield by describing the state of the art in concentric tube robot research. We begin with current and proposed applications for these robots and then trace their origins (some aspects of which date back to 1985), before proceeding to describe the state of the art in terms of modeling, control, sensing, and design. The paper concludes with forward-looking perspectives, noting that concentric tube robots provide rich opportunities for further research, yet simultaneously appear poised to become viable commercial devices in the near future.

Hunter B. Gilbert, D. Caleb Rucker, Robert J. Webster III

A Framework for Real-Time Multi-Contact Multi-Body Dynamic Simulation

In this paper we propose a unified framework for the real-time dynamic simulation and contact resolution of rigid articulated bodies. This work builds on previous developments in the field of dynamic simulation, collision detection, contact resolution, and operational space control. However, the key to efficiency and real-time performance is a new parallel implementation of our collision detection and contact resolution algorithm which decomposes the problem into tasks that can be concurrently executed. Finally, the results and accuracy of our simulation models are compared for the first time against recorded motions of real articulated bodies colliding on a frictionless air floating table.

François Conti, Oussama Khatib

Intelligence and Learning


Constructive Developmental Science: A Trans-Disciplinary Approach Toward the Fundamentals of Human Cognitive Development and Its Disorders, Centered Around Fetus Simulation

How does human mind develop? What causes developmental disorders? Recent studies suggest the importance of the fetal period in human development. However, study of human fetuses is strongly constrained by technical and ethical difficulties. This project aims at understanding the principles of human development by analyzing and modeling it from the fetal period. Integrating robotics, medicine, psychology, neuroscience, and Tohjisha-kenkyu (first-person view research of developmental disorders), we establish a new trans-disciplinary research field called Constructive Developmental Science. Its contributions include a new understanding of human development and its disorders, comprehensive diagnostic methodologies, and truly appropriate assistive technology.

Yasuo Kuniyoshi

Personalizing Intelligent Systems and Robots with Human Motion Data

According to Merhabian, more than $$90\,\%$$90% of human-human communication is non-verbal when expressing affects and attitudes. Further studies have shown that a large proportion of non-verbal communication can be attributed to posture and to gesture. They communicate information about action: intent, meaning, as well as information about internal states such as affects. Emotional understanding is a key for satisfying and successful interaction between two or more humans, it must also be true for human-robot interaction. In this paper we explore the importance of non verbal information and communication, typically motion data, and how it can be used to develop and to personalize intelligent systems and robots. First, we present and discuss our findings on the strong correlation between what humans feel during an unannounced interaction with a humanoid robot and their movements and attitudes. Then, we propose a framework that uses not only the kinematics information of movements but also the dynamics. We use the direct measure of the dynamics when available. If not we propose to compute the dynamics from the kinematics, and use it to understand human motions. Finally, we discuss some developments and concrete applications in the field of health care and HRI.

Gentiane Venture, Ritta Baddoura, Yuta Kawashima, Noritaka Kawashima, Takumi Yabuki

Beyond Geometric Path Planning: Learning Context-Driven Trajectory Preferences via Sub-optimal Feedback

We consider the problem of learning preferences over trajectories for mobile manipulators such as personal robots and assembly line robots. The preferences we learn are more intricate than those arising from simple geometric constraints on robot’s trajectory, such as distance of the robot from human etc. Our preferences are rather governed by the surrounding context of various objects and human interactions in the environment. Such preferences makes the problem challenging because the criterion of defining a good trajectory now varies with the task, with the environment and across the users. Furthermore, demonstrating optimal trajectories (e.g., learning from expert’s demonstrations) is often challenging and non-intuitive on high degrees of freedom manipulators. In this work, we propose an approach that requires a non-expert user to only incrementally improve the trajectory currently proposed by the robot. We implement our algorithm on two high degree-of-freedom robots, PR2 and Baxter, and present three intuitive mechanisms for providing such incremental feedback. In our experimental evaluation we consider two context rich settings—household chores and grocery store checkout—and show that users are able to train the robot with just a few feedbacks (taking only a few minutes). Despite receiving sub-optimal feedback from non-expert users, our algorithm enjoys theoretical bounds on regret that match the asymptotic rates of optimal trajectory algorithms.

Ashesh Jain, Shikhar Sharma, Ashutosh Saxena

Learning from Demonstrations Through the Use of Non-rigid Registration

We consider the problem of teaching robots by demonstration how to perform manipulation tasks, in which the geometry (including size, shape, and pose) of the relevant objects varies from trial to trial. We present a method, which we call trajectory transfer, for adapting a demonstrated trajectory from the geometry at training time to the geometry at test time. Trajectory transfer is based on non-rigid registration, which computes a smooth transformation from the training scene onto the testing scene. We then show how to perform a multi-step task by repeatedly looking up the nearest demonstration and then applying trajectory transfer. As our main experimental validation, we enable a PR2 robot to autonomously tie five different types of knots in rope.

John Schulman, Jonathan Ho, Cameron Lee, Pieter Abbeel



Robust Contact Generation for Robot Simulation with Unstructured Meshes

This paper presents a numerically stable method for rigid body simulation of unstructured meshes undergoing forceful contact, such as in robot locomotion and manipulation. The key contribution is a new contact generation method that treats the geometry as having a thin virtual boundary layer around the underlying meshes. Unlike existing methods, it produces contact estimates that are stable with respect to small displacements, which helps avoid jitter or divergence in the simulator caused by oscillatory discontinuities. Its advantages are particularly apparent on non-watertight meshes and can easily simulate interaction with partially-sensed and noisy objects, such as those that emerge from low-cost 3D scanners. The simulator is tested on a variety of robot locomotion and manipulation examples, and results closely match theoretical predictions and experimental data.

Kris Hauser

Manifold Representations for State Estimation in Contact Manipulation

We investigate the problem of using contact sensors to estimate the configuration of an object during manipulation. Contact sensing is very discriminative by nature and, therefore, the set of object configurations that activate a sensor constitutes a lower-dimensional manifold in the configuration space of the object. This causes conventional state estimation methods, such as particle filters, to perform poorly during periods of contact. The manifold particle filter addresses this problem by sampling particles directly from the contact manifold. When it exists, we can sample these particles from an analytic representation of the contact manifold. We present two alternative sample-based contact manifold representations that make no assumptions about the object-hand geometry: rejection sampling and trajectory rollouts. We discuss theoretical considerations behind these three representations and compare their performance in a suite of simulation experiments. We show that all three representations enable the manifold particle filter to outperform the conventional particle filter. Additionally, we show that the trajectory rollout representation performs similarly to the analytic method despite the rollout method’s relative simplicity.

Michael C. Koval, Nancy S. Pollard, Siddhartha S. Srinivasa

Exploitation of Environmental Constraints in Human and Robotic Grasping

We investigate the premise that robust grasping performance is enabled by exploiting constraints present in the environment. These constraints, leveraged through motion in contact, counteract uncertainty in state variables relevant to grasp success. Given this premise, grasping becomes a process of successive exploitation of environmental constraints, until a successful grasp has been established. We present support for this view by analyzing human grasp behavior and by showing robust robotic grasping based on constraint-exploiting grasp strategies. Furthermore, we show that it is possible to design robotic hands with inherent capabilities for the exploitation of environmental constraints.

Raphael Deimel, Clemens Eppner, José Álvarez-Ruiz, Marianne Maertens, Oliver Brock

Restraining Objects with Curved Effectors and Its Application to Whole-Arm Grasping

This paper develops the theory and algorithms for immobilizing/caging polyhedral objects using curved (for example, planar, cylindrical, or spherical) effectors, in contrast to customary point effectors. We show that it is possible to immobilize all polyhedral objects with three effectors with possibly nonzero curvature, with finite extent. We further discuss how to cage the objects and obtain a stable grasp from such a cage. The theory can also be applied to immobilize/cage polygonal objects on the plane. As one application of the theory, we address the problem of whole-arm grasping with robot arms.

Jungwon Seo, Mark Yim, Vijay Kumar



Data Association for Semantic World Modeling from Partial Views

Autonomous mobile-manipulation robots need to sense and interact with objects to accomplish high-level tasks such as preparing meals and searching for objects. To achieve such tasks, robots need semantic world models, defined as object-based representations of the world involving task-level attributes. In this work, we address the problem of estimating world models from semantic perception modules that provide noisy observations of attributes. Because attribute detections are sparse, ambiguous, and are aggregated across different viewpoints, it is unclear which attribute measurements are produced by the same object, so data association issues are prevalent. We present novel clustering-based approaches to this problem, which are more efficient and require less severe approximations compared to existing tracking-based approaches. These approaches are applied to data containing object type-and-pose detections from multiple viewpoints, and demonstrate comparable quality to the existing approach using a fraction of the computation time.

Lawson L. S. Wong, Leslie Pack Kaelbling, Tomás Lozano-Pérez

Driven Learning for Driving: How Introspection Improves Semantic Mapping

This paper explores the suitability of commonly employed classification methods to action-selection tasks in robotics, and argues that a classifier’s introspective capacity is a vital but as yet largely under-appreciated attribute. As illustration we propose an active learning framework for semantic mapping in mobile robotics and demonstrate it in the context of autonomous driving. In this framework, data are selected for label disambiguation by a human supervisor using uncertainty sampling. Intuitively, an introspective classification framework—i.e. one which moderates its predictions by an estimate of how well it is placed to make a call in a particular situation—is particularly well suited to this task. To achieve an efficient implementation we extend the notion of introspection to a particular sparse Gaussian Process Classifier, the Informative Vector Machine (IVM). Furthermore, we leverage the information-theoretic nature of the IVM to formulate a principled mechanism for forgetting stale data, thereby bounding memory use and resulting in a truly life-long learning system. Our evaluation on a publicly available dataset shows that an introspective active learner asks more informative questions compared to a more traditional non-introspective approach like a Support Vector Machine (SVM) and in so doing, outperforms the SVM in terms of learning rate while retaining efficiency for practical use.

Rudolph Triebel, Hugo Grimmett, Rohan Paul, Ingmar Posner

RatSLAM: Using Models of Rodent Hippocampus for Robot Navigation and Beyond

We describe recent biologically-inspired mapping research incorporating brain-based multi-sensor fusion and calibration processes and a new multi-scale, homogeneous mapping framework. We also review the interdisciplinary approach to the development of the RatSLAM robot mapping and navigation system over the past decade and discuss the insights gained from combining pragmatic modelling of biological processes with attempts to close the loop back to biology. Our aim is to encourage the pursuit of truly interdisciplinary approaches to robotics research by providing successful case studies.

Michael Milford, Adam Jacobson, Zetao Chen, Gordon Wyeth

Into Darkness: Visual Navigation Based on a Lidar-Intensity-Image Pipeline

Visual navigation of mobile robots has become a core capability that enables many interesting applications from planetary exploration to self-driving cars. While systems built on passive cameras have been shown to be robust in well-lit scenes, they cannot handle the range of conditions associated with a full diurnal cycle. Lidar, which is fairly invariant to ambient lighting conditions, offers one possible remedy to this problem. In this paper, we describe a visual navigation pipeline that exploits lidar’s ability to measure both range and intensity (a.k.a., reflectance) information. In particular, we use lidar intensity images (from a scanning-laser rangefinder) to carry out tasks such as visual odometry (VO) and visual teach and repeat (VT&R) in realtime, from full-light to full-dark conditions. This lighting invariance comes at the price of coping with motion distortion, owing to the scanning-while-moving nature of laser-based imagers. We present our results and lessons learned from the last few years of research in this area.

Timothy D. Barfoot, Colin McManus, Sean Anderson, Hang Dong, Erik Beerepoot, Chi Hay Tong, Paul Furgale, Jonathan D. Gammell, John Enright

Automatic Differentiation on Differentiable Manifolds as a Tool for Robotics

Automatic differentiation (AD) is a useful tool for computing Jacobians of functions needed in estimation and control algorithms. However, for many interesting problems in robotics, state variables live on a differentiable manifold. The most common example are robot orientations that are elements of the Lie group SO(3). This causes problems for AD algorithms that only consider differentiation at the scalar level. Jacobians produced by scalar AD are correct, but scalar-focused methods are unable to apply simplifications based on the structure of the specific manifold. In this paper we extend the theory of AD to encompass handling of differentiable manifolds and provide a C++ library that exploits strong typing and expression templates for fast, easy-to-use Jacobian evaluation. This method has a number of benefits over scalar AD. First, it allows the exploitation of algebraic simplifications that make Jacobian evaluations more efficient than their scalar counterparts. Second, strong typing reduces the likelihood of programming errors arising from misinterpretation that are possible when using simple arrays of scalars. To the best of our knowledge, this is the first work to consider the structure of differentiable manifolds directly in AD.

Hannes Sommer, Cédric Pradalier, Paul Furgale

Minimal Solutions for Pose Estimation of a Multi-Camera System

In this paper, we propose a novel formulation to solve the pose estimation problem of a calibrated multi-camera system. The non-central rays that pass through the 3D world points and multi-camera system are elegantly represented as Plücker lines. This allows us to solve for the depth of the points along the Plücker lines with a minimal set of 3-point correspondences. We show that the minimal solution for the depth of the points along the Plücker lines is an 8 degree polynomial that gives up to 8 real solutions. The coordinates of the 3D world points in the multi-camera frame are computed from the known depths. Consequently, the pose of the multi-camera system, i.e. the rigid transformation between the world and multi-camera frames can be obtained from absolute orientation. We also derive a closed-form minimal solution for the absolute orientation. This removes the need for the computationally expensive Singular Value Decompositions (SVD) during the evaluations of the possible solutions for the depths. We identify the correct solution and do robust estimation with RANSAC. Finally, the solution is further refined by including all the inlier correspondences in a non-linear refinement step. We verify our approach by showing comparisons with other existing approaches and results from large-scale real-world datasets.

Gim Hee Lee, Bo Li, Marc Pollefeys, Friedrich Fraundorfer

Recursive Inference for Prediction of Objects in Urban Environments

Future advancements in robotic navigation and mapping rest to a large extent on robust, efficient and more advanced semantic understanding of the surrounding environment. The existing semantic mapping approaches typically consider small number of semantic categories, require complex inference or large number of training examples to achieve desirable performance. In the proposed work we present an efficient approach for predicting locations of generic objects in urban environments by means of semantic segmentation of a video into object and non-object categories. We exploit widely available exemplars of non-object categories (such as road, buildings, vegetation) and use geometric cues which are indicative of the presence of object boundaries to gather the evidence about objects regardless of their category. We formulate the object/non-object semantic segmentation problem in the Conditional Random Field framework, where the structure of the graph is induced by a minimum spanning tree computed over a 3D point cloud, yielding an efficient algorithm for an exact inference. The chosen 3D representation naturally lends itself for on-line recursive belief updates with a simple soft data association mechanism. We carry out extensive experiments on videos of urban environments acquired by a moving vehicle and show quantitatively and qualitatively the benefits of our proposal.

Cesar Cadena, Jana Košecká

A New Approach to Model-Free Tracking with 2D Lidar

This paper presents a unified and model-free framework for the detection and tracking of dynamic objects with 2D laser range finders in an autonomous driving scenario. A novel state formulation is proposed that captures joint estimates of the sensor pose, a local static background and dynamic states of moving objects. In addition, we contribute a new hierarchical data association algorithm to associate raw laser measurements to observable states, and within which, a new variant of the Joint Compatibility Branch and Bound (JCBB) algorithm is introduced for problems with large numbers of measurements. The system is calibrated systematically on 7.5K labeled object examples and evaluated on 6K test cases, and is shown to greatly outperform an existing industry standard targeted at the same problem domain.

Dominic Zeng Wang, Ingmar Posner, Paul Newman



Task-Oriented Grasp Planning Based on Disturbance Distribution

One difficulty of task-oriented grasp planning is task modeling. In this paper, a manipulation task was modeled by building a non-parametric statistical distribution model from disturbance data captured during demonstrations. This paper proposes a task-oriented grasp quality criterion based on distribution of task disturbance and uses the criterion to search for a grasp that covers the most significant part of the disturbance distribution. To reduce the computational complexity of the search in a high-dimensional robotic hand configuration space, as well as to avoid a correspondence problem, the candidate grasps are computed from a reduced configuration space that is confined by a set of given thumb placements and thumb directions. The proposed approach has been validated with a Barrett hand and a Shadow hand on several objects in simulation. The resulting grasps in the evaluation generated by our approach increase the coverage of frequently-occurring disturbance rather than the coverage of a large area with a scattered distribution.

Yun Lin, Yu Sun

Towards Planning in Generalized Belief Space

We investigate the problem of planning under uncertainty, which is of interest in several robotic applications, ranging from autonomous navigation to manipulation. Recent effort from the research community has been devoted to design planning approaches working in a continuous domain, relaxing the assumption that the controls belong to a finite set. In this case robot policy is computed from the current robot belief (planning in belief space), while the environment in which the robot moves is usually assumed to be known or partially known. We contribute to this branch of the literature by relaxing the assumption of known environment; for this purpose we introduce the concept of generalized belief space (GBS), in which the robot maintains a joint belief over its state and the state of the environment. We use GBS within a Model Predictive Control (MPC) scheme; our formulation is valid for general cost functions and incorporates a dual-layer optimization: the outer layer computes the best control action, while the inner layer computes the generalized belief given the action. The resulting approach does not require prior knowledge of the environment and does not assume maximum likelihood observations. We also present an application to a specific family of cost functions and we elucidate on the theoretical derivation with numerical examples.

Vadim Indelman, Luca Carlone, Frank Dellaert

An Online POMDP Solver for Uncertainty Planning in Dynamic Environment

Motion planning under uncertainty is important for reliable robot operations in uncertain and dynamic environments. Partially Observable Markov Decision Process (POMDP) is a general and systematic framework for motion planning under uncertainty. To cope with dynamic environment well, we often need to modify the POMDP model during runtime. However, despite recent tremendous advances in POMDP planning, most solvers are not fast enough to generate a good solution when the POMDP model changes during runtime. Recent progress in online POMDP solvers have shown promising results. However, most online solvers are based on replanning, which recompute a solution from scratch at each step, discarding any solution that has been computed so far, and hence wasting valuable computational resources. In this paper, we propose a new online POMDP solver, called Adaptive Belief Tree (ABT), that can reuse and improve existing solution, and update the solution as needed whenever the POMDP model changes. Given enough time, ABT converges to the optimal solution of the current POMDP model in probability. Preliminary results on three distinct robotics tasks in dynamic environments are promising. In all test scenarios, ABT generates similar or better solutions faster than the fastest online POMDP solver today; using an average of less than 50 ms of computation time per step.

Hanna Kurniawati, Vinay Yadav

An Enzyme-Inspired Approach to Stochastic Allocation of Robotic Swarms Around Boundaries

This work presents a novel control approach for allocating a robotic swarm among boundaries. It represents the first step toward developing a methodology for encounter-based swarm allocation that incorporates rigorously characterized spatial effects in the system without requiring analytical expressions for encounter rates. Our approach utilizes a macroscopic model of the swarm population dynamics to design stochastic robot control policies that result in target allocations of robots to the boundaries of regions of different types. The control policies use only local information and have provable guarantees on the collective swarm behavior. We analytically derive the relationship between the stochastic control policies and target allocations for a scenario in which circular robots avoid collisions with each other, bind to boundaries of disk-shaped regions, and command bound robots to unbind. We validate this relationship in simulation and show that it is robust to environmental changes, such as a change in the number or size of robots and disks.

Theodore P. Pavlic, Sean Wilson, Ganesh P. Kumar, Spring Berman

Polynomial Trajectory Planning for Aggressive Quadrotor Flight in Dense Indoor Environments

We explore the challenges of planning trajectories for quadrotors through cluttered indoor environments. We extend the existing work on polynomial trajectory generation by presenting a method of jointly optimizing polynomial path segments in an unconstrained quadratic program that is numerically stable for high-order polynomials and large numbers of segments, and is easily formulated for efficient sparse computation. We also present a technique for automatically selecting the amount of time allocated to each segment, and hence the quadrotor speeds along the path, as a function of a single parameter determining aggressiveness, subject to actuator constraints. The use of polynomial trajectories, coupled with the differentially flat representation of the quadrotor, eliminates the need for computationally intensive sampling and simulation in the high dimensional state space of the vehicle during motion planning. Our approach generates high-quality trajecrtories much faster than purely sampling-based optimal kinodynamic planning methods, but sacrifices the guarantee of asymptotic convergence to the global optimum that those methods provide. We demonstrate the performance of our algorithm by efficiently generating trajectories through challenging indoor spaces and successfully traversing them at speeds up to 8 m/s. A demonstration of our algorithm and flight performance is available at:

Charles Richter, Adam Bry, Nicholas Roy

Fast Marching Trees: A Fast Marching Sampling-Based Method for Optimal Motion Planning in Many Dimensions

In this paper we present a novel probabilistic sampling-based motion planning algorithm called the Fast Marching Tree algorithm (FMT$$^*$$∗). The algorithm is specifically aimed at solving complex motion planning problems in high-dimensional configuration spaces. This algorithm is proven to be asymptotically optimal and is shown to converge to an optimal solution faster than its state-of-the-art counterparts, chiefly PRM$$^*$$∗ and RRT$$^*$$∗. An additional advantage of $$\text {FMT}^*$$FMT∗ is that it builds and maintains paths in a tree-like structure (especially useful for planning under differential constraints). The $$\text {FMT}^*$$FMT∗ algorithm essentially performs a “lazy” dynamic programming recursion on a set of probabilistically-drawn samples to grow a tree of paths, which moves steadily outward in cost-to-come space. As such, this algorithm combines features of both single-query algorithms (chiefly RRT) and multiple-query algorithms (chiefly PRM), and is conceptually related to the Fast Marching Method for the solution of eikonal equations. As a departure from previous analysis approaches that are based on the notion of almost sure convergence, the $$\text {FMT}^*$$FMT∗ algorithm is analyzed under the notion of convergence in probability: the extra mathematical flexibility of this approach allows for significant algorithmic advantages and provides convergence rate bounds—a first in the field of optimal sampling-based motion planning. Numerical experiments over a range of dimensions and obstacle configurations confirm our theoretical and heuristic arguments by showing that FMT$$^*$$∗, for a given execution time, returns substantially better solutions than either PRM$$^*$$∗ or RRT$$^*$$∗, especially in high-dimensional configuration spaces and in scenarios where collision checking is expensive.

Lucas Janson, Marco Pavone

Safe Motion Planning for Imprecise Robotic Manipulators by Minimizing Probability of Collision

Robotic manipulators designed for home assistance and new surgical procedures often have significant uncertainty in their actuation due to compliance requirements, cost constraints, and size limits. We introduce a new integrated motion planning and control algorithm for robotic manipulators that makes safety a priority by explicitly considering the probability of unwanted collisions. We first present a fast method for estimating the probability of collision of a motion plan for a robotic manipulator under the assumptions of Gaussian motion and sensing uncertainty. Our approach quickly computes distances to obstacles in the workspace and appropriately transforms this information into the configuration space using a Newton method to estimate the most relevant collision points in configuration space. We then present a sampling-based motion planner based on executing multiple independent rapidly exploring random trees that returns a plan that, under reasonable assumptions, asymptotically converges to a plan that minimizes the estimated collision probability. We demonstrate the speed and safety of our plans in simulation for (1) a 3-D manipulator with 6 DOF, and (2) a concentric tube robot, a tentacle-like robot designed for surgical applications.

Wen Sun, Luis G. Torres, Jur van den Berg, Ron Alterovitz
Weitere Informationen