ICRA 2015 Papers

Existing control techniques for optical tweezers utilize a single focused laser beam to directly trap and manipulate a target cell. However, a typical force generated by an optical trap is extremely small (few pico-newtons) and therefore it is not sufficient to manipulate a larger cell or object. The optical trap is also sensitive to the shape of the biological cell and the refractive index. Therefore, current automatic control techniques for optical tweezers cannot be used to manipulate a large cell or cell with irregular shape. In addition, excessive irradiation of the laser beam to the cell may also cause photodamage of even lead to death of the cell. In this paper, we propose a robotic control technique for optical tweezers to achieve automated manipulation of cell, which is beyond the capability of a single optical trap. First, multiple laser beams are generated, and each laser beam is used to trap one micro-particle to create a formation around the target cell to hold it. Then the target cell is manipulated to a desired position by controlling the motorized stage. The proposed control technique is particularly suitable for automated manipulation of sensitive biological cells, cells with large size or cells of irregular shape. Rigorous mathematical formulations have been developed to analyze the control system for automated cell manipulation. Experimental results are presented to illustrate the performance of the proposed controller.

For using industrial robots in applications where the robot physically interacts with the environment, such as assembly, force control is usually needed. A force sensor may, however, be expensive and add mass to the system. An alternative is therefore to estimate the external force using the motor torques. This paper considers the problem of force estimation for the case when the robot is not moving, where the Coulomb friction constitutes a fundamental difficulty. A dithering feedforward torque is used to decrease the Coulomb friction uncertainty, and hence improve the force estimation accuracy when the robot is not moving. The method is validated experimentally through an implementation on an industrial robot. A lead-through scenario is also presented.

This paper proposes a framework called Episodic memory-driving Markov decision processes (EM-MDPs) for incremental self-learning of robotic experience and cognitive behavior control under uncertainty. The framework simulates the organization process of episodic memory by introducing the neuron stimulation mechanism. Firstly, episode model is built, and the activation and stimulation mechanism of state neurons is proposed based on cognitive neuroscience. Secondly, episodic self-learning is also proposed by utilizing sparse distributed memory (SDM) through Hebbian rules, to realize memory real-time storage, incremental accumulation and integration. Finally, a robotic cognitive behavior control approach is established. Neuron synaptic potential is introduced for event localization. Robot can evaluate the past events sequence, predict the current state and plan the desired behavior. Two main challenges in robot behavior control under uncertainty are addressed in the paper: high computational complexity and perceptual aliasing. The proposed system is evaluated in several real life environments for mobile robot. The applicability and the usefulness of the developed method are validated by the results obtained.

We present an extension, TeleoR, of Nilsson's Teleo-Reactive (TR) rule based robotic agent programming language[22]. For both languages programs essentially comprise sequences of Guard ~>Action rules grouped into parameterised procedures. The Guard is a deductive query to a set of rapidly changing percept facts generated from the most recent sensor values. For TR, the Action is either a tuple of primitive robotic actions, to be executed in parallel, or a single call to a program procedure, which can be a recursive call, or a BeliefStoreupdate. TeleoR has extra forms of action. The procedures encode goal (teleo) directed reactive task and sub-task behaviours of robotic agents. TR/TeleoR programs are robust and opportunity grabbing, and so are well suited to human/robot or robot/robot co-operative tasks requiring flexible behaviour. TeleoR, extends TR in having: types and higher order features; extra forms of action rules that temporarily inhibit other rules in the same procedure; repeatable sequences of time capped actions; wait/repeat re-start of failed actions; belief store update and message send actions linked with any rule; a flexibly typed higher order LP/FP programming language for BeliefStore inference; support for the high level programming of multi-tasking agents that interleave the use of subsets of a set of independent robotic resources. All the extensions were driven by application needs. The use of QuLog enables us to guarantee by compiler analysis that all guarded rule actions will be fully instantiated and correctly typed when sent to a robot, perhaps via a ROS interface. The focus of this paper is on the extensions for single task communicating robotic agents. Our goal was to extend TR without losing the elegance and simplicity of Nilsson's language. We also wanted to be able to give the extended language a formally defined operational semantics, building upon one we had given for TR. The extensions, their semantics, and their implementation were dev... Show More

Biologically inspired robots perform many interesting and useful behaviors, but to effectively emulate their biological counterparts, robots often need to possess many degrees of freedom, complicating their mechanical design and making it difficult to apply standard control and motion planning strategies. To address this complexity, the robotics community has derived low-dimensional parameter-based controllers that naturally coordinate many degrees of freedom such as the serpenoid curves used to control snake robots. Controllers utilizing this parameterization for snake robots have been able to induce behaviors similar to that of the robots' biological counterparts. A similar concept, called a control template, is used in the study of animal movements. However, much of the prior work on control templates has been limited to in-plane motion. In this work, we extend the usage of control templates to three dimensions to both better model and understand biology, as well as to help us gain better intuition into how we can use pre-existing control paradigms to create new behaviors for biologically inspired robots.

As service robots become more and more capable of performing useful tasks for us, there is a growing need to teach robots how we expect them to carry out these tasks. However, learning our preferences is a nontrivial problem, as many of them stem from a variety of factors including personal taste, cultural background, or common sense. Obviously, such factors are hard to formulate or model a priori. In this paper, we present a solution for tidying up objects in containers, e.g., shelves or boxes, by following user preferences. We learn the user preferences using collaborative filtering based on crowdsourced and mined data. First, we predict pairwise object preferences of the user. Then, we subdivide the objects in containers by modeling a spectral clustering problem. Our solution is easy to update, does not require complex modeling, and improves with the amount of user data. We evaluate our approach using crowdsoucing data from over 1,200 users and demonstrate its effectiveness for two tidy-up scenarios. Additionally, we show that a real robot can reliably predict user preferences using our approach.

In this paper we present for the first time a complete symbolic navigation system that performs goal-directed exploration to unfamiliar environments on a physical robot. We introduce a novel construct called the abstract map to link provided symbolic spatial information with observed symbolic information and actual places in the real world. Symbolic information is observed using a text recognition system that has been developed specifically for the application of reading door labels. In the study described in this paper, the robot was provided with a floor plan and a destination. The destination was specified by a room number, used both in the floor plan and on the door to the room. The robot autonomously navigated to the destination using its text recognition, abstract map, mapping, and path planning systems. The robot used the symbolic navigation system to determine an efficient path to the destination, and reached the goal in two different real-world environments. Simulation results show that the system reduces the time required to navigate to a goal when compared to random exploration.

Robot Programming by Demonstration (PbD) allows users to program a robot by demonstrating the desired behavior. Providing these demonstrations typically involves moving the robot through a sequence of states, often by physically manipulating it. This requires users to be co-located with the robot and have the physical ability to manipulate it. In this paper, we present a natural language based interface for PbD that removes these requirements and enables hands-free programming. We focus on programming object manipulation actions—our key insight is that such actions can be decomposed into known types of manipulator movements that are naturally described using spatial language; e.g., object reference expressions and prepositions. Our method takes a natural language command and the current world state to infer the intended movement command and its parametrization. We implement this method on a two-armed mobile manipulator and demonstrate the different types of manipulation actions that can be programmed with it. We compare it to a kinesthetic PbD interface and we demonstrate our method's ability to deal with incomplete language.

We present RoboSherlock, an open source software framework for implementing perception systems for robots performing human-scale everyday manipulation tasks. In RoboSherlock, perception and interpretation of realistic scenes is formulated as an unstructured information management (UIM) problem. The application of the UIM principle supports the implementation of perception systems that can answer task-relevant queries about objects in a scene, boost object recognition performance by combining the strengths of multiple perception algorithms, support knowledge-enabled reasoning about objects and enable automatic and knowledge-driven generation of processing pipelines. We demonstrate the potential of the proposed framework by three feasibility studies of systems for real-world scene perception that have been built on top of RoboSherlock.

General-purpose robots can perform a range of useful tasks in human environments; however, programming them to robustly function in all possible environments that they might encounter is unfeasible. Instead, our research aims to develop robots that can be programmed by its end-users in their context of use, so that the robot needs to robustly function in only one particular environment. This requires intuitive ways in which end-users can program their robot. To that end, this paper contributes a flow-based visual programming language, called RoboFlow, that allows programming of generalizable mobile manipulation tasks. RoboFlow is designed to (i) ensure a robust low-level implementation of program procedures on a mobile manipulator, and (ii) restrict the high-level programming as much as possible to avoid user errors while enabling expressive programs that involve branching, looping, and nesting. We present an implementation of RoboFlow on a PR2 mobile manipulator and demonstrate the generalizability and error handling properties of RoboFlow programs on everyday mobile manipulation tasks in human environments.

A robot companion should adapt to its user's needs by learning to perform new tasks. In this paper, we present a robot playmate that learns and adapts to tasks chosen by the child on a touchscreen tablet. We aim to solve the task learning problem using an experience-based learning framework that stores human demonstrations as task instances. These instances are retrieved when confronted with a similar task in which the system generates predictions of task behaviors based on prior actions. In order to automate the processes of instance encoding, acquisition, and retrieval, we have developed a framework that gathers task knowledge through interaction with human teachers. This approach, further referred to as interactive instance-based learning (IIBL), utilizes limited information available to the robot to generate similarity metrics for retrieving instances. In this paper, we focus on introducing and evaluating a new hybrid IIBL framework using sensitivity analysis with artificial neural networks and discuss its advantage over methods using k-NNs and linear regression in retrieving instances.

This paper proposes a generalized algorithm that enables mobile agents to meet in a bounded region based only on bearing information of other agents within their vicinity. Each agent repeatedly employs a stop-and-go strategy consisting of the following three actions: (1) Estimate the bearing of agents in its vicinity, (2) compute a target point based on the estimates, and (3) move to that target point. The motivation and case study example is a modular robot, the Distributed Flight Array, which we employ to validate the proposed algorithm.

Robotic needle steering systems for minimally invasive medical procedures require complementary medical imaging systems to track the needles in real time. Ultrasound is a promising imaging modality because it offers relatively low-cost, real-time imaging of the needle. Previous methods applied vibration to the base of the needle using a voice coil actuator, in order to make the needle visible in power Doppler ultrasound. We propose a new method for needle tip vibration, using electromagnetic actuation of small permanent magnets placed inside the needle to improve needle tip visibility in power Doppler imaging. Robotic needle insertion experiments using artificial tissue and ex vivo porcine liver showed that the electromagnetic tip vibration method can generate a stronger Doppler response compared to the previous base vibration method, resulting in better imaging at greater needle depth in tissue. It also eliminates previous issues with vibration damping along the shaft of the needle.

Practical mapping and navigation solutions for large indoor environments continue to rely on relatively expensive range scanners, because of their accuracy, range and field of view. Microsoft Kinect on the other hand is inexpensive, is easy to use and has high resolution, but suffers from high noise, shorter range and a limiting field of view. We present a mapping and navigation system that uses the Microsoft Kinect sensor as the sole source of range data and achieves performance comparable to state-of-the-art LIDAR-based systems. We show how we circumvent the main limitations of Kinect to generate usable 2D maps of relatively large spaces and to enable robust navigation in changing and dynamic environments. We use the Benchmark for Robotic Indoor Navigation (BRIN) to quantify and validate the performance of our system.

Replacing aircraft hydraulic actuators with ElectroMechanical Actuators (EMAs) is seen as a necessary step towards the development of safer and more efficient aircraft with minimal environmental impact. EMAs are commonly used in a wide range of applications as they have improved efficiency over hydraulic actuators. However, in their current form, EMAs do not comply with the stringent weight and reliability requirements of primary flight control applications. This paper describes an Active Torque Distribution (ATD) system that could meet flight control reliability requirements, be lightweight, and increase the dynamic performance of current flight controls. The proposed ATD system uses magneto-rheological clutches to modulate the torque transferred from a centralized velocity source to various control surfaces. The fundamental advantage of this approach is found in the fluidic nature of magneto-rheological clutch interfaces which transmit torque without any solid mechanical contact thus preventing wear and failure causing mechanical seizure. The analytical study used to scale an ATD system and an analogue EMA system based on a representative flight control application shows that the proposed ATD system is 42% lighter than a system composed of EMAs and has a weight similar to that of current hydraulic actuators. Furthermore, this study demonstrates that ATD systems add ~100x less inertia to flight surfaces than EMAs allowing implementation of load alleviation control algorithms requiring high frequency motion (~30 Hz). A proof-of-concept clutch and controller confirm the analytical predictions and validate the possibility of using magneto-rheological clutches for flight control actuation.

Deformable objects such as garments are highly unstructured, making them difficult to recognize and manipulate. In this paper, we propose a novel method to teach a two-arm robot to efficiently track the states of a garment from an unknown state to a known state by iterative regrasping. The problem is formulated as a constrained weighted evaluation metric for evaluating the two desired grasping points during regrasping, which can also be used for a convergence criterion The result is then adopted as an estimation to initialize a regrasping, which is then considered as a new state for evaluation. The process stops when the predicted thin shell conclusively agrees with reconstruction. We show experimental results for regrasping a number of different garments including sweater, knitwear, pants, and leggings, etc.

This paper presents a method to align a virtual plane with a real plane in an augmented reality environment. The method addresses the challenge of configuring a planar virtual fixture during the teleoperation of a robot in a visually challenging environment. The method assists the operator to manually align the virtual plane with the real plane by providing visual cues by dynamically texturing the virtual plane. Following the manual initialization a robust evaluation of the optical flow is used to refine the alignment of the virtual plane. The system combines a stereo camera with a stereo display and generates visual distortions in the displayed images to indicate a misalignment of the virtual plane with the real one. The distortion is computed by displaying warped images from the left camera in the right display and vice versa. Results demonstrate that the method improves the accuracy of the alignment with a real plane that is covered with reflective multi-layer insulation.

This paper proposes an optimization method for dual-arm robots, inspired by a study on human asymmetric bimanual action called Guiard's principle which states that when humans perform asymmetric bimanual tasks, the right hand (as the lateral preference) performs a fine motion (and force) resolution, while the left hand performs a coarse resolution. To effectively transfer the human bimanual-task knowledge to dual-arm robots, we proposed a cost function, based on task-compatibility index, which is used to set the desired motion and force resolution for each end-effector of the dual-arm according to its role in the bimanual action. Thus the right-arm posture is optimized such that the tool attached to its end-effector can exert fine motion and force. And the left-arm posture is optimized to assume a strong and dynamic structural support for the right arm action. We experimentally compare the proposed cost function against previous methods for dual-arm robots using two six degrees-of-freedom torque-controlled manipulators. The control performance, when mimicking Guiard's principle shows considerably better results.

This paper proposes the emulation of the socalled kingpin mechanism in order to reduce the off-tracking effects exhibited by the standard and generalized n-trailer systems. In turn, both systems are emulated by a group of differential-drive mobile robots trough time-varying relative position vectors using the leader-follower scheme. A control law is proposed which involves the position of the virtual sliding kingpin mechanism. It is shown that when the virtual kingpin mechanism is used in the control law, the off-tracking effects are significantly reduced compared to when this virtual mechanism is not used. Numerical simulations and real-time experiments show the performance of the proposed control law.