ICRA 2017 Papers

In this paper we propose a probabilistic approach for detecting and classifying urban road intersections from a moving vehicle. The approach is based on images from an onboard stereo rig; it relies on the detection of the road ground plane on one side, and on a pixel-level classification of the road on the other. The two processing pipelines are then integrated and the parameters of the road components, i.e., the intersection geometry, are inferred. As opposed to other state-of-the-art offline methods, which require processing of the whole video sequence, our approach integrates the image data by means of an online procedure. The experiments have been performed on well-known KITTI datasets, allowing for future comparisons.

A robot mediator can enhance the quality of patient care in a health care context. Patients with Parkinson's disease can experience difficulties in precisely expressing their emotions due to the loss of control of their facial musculature, leading to their stigmatization by caregivers. To remedy this challenge, a robot mediator can be inserted into a patient-caregiver relationship. In this context, it is essential to handle the ethical issues of neglect to ensure human dignity. In an earlier paper [19], we proposed an intervening ethical governor (IEG) model, which enables a robot to ethically intervene in a situation where patients or caregivers go across accepted ethical boundaries. In this paper, we show how the IEG model can be implemented and applied in a real robotics system. In addition, by conducting interviews with the target population (adults 60 years of age or older), we evaluate the current intervention rules in the model, discuss potential improvements to the model, and consider uses of the model in real clinical contexts.

Widely-used methods for toolpath planning in fused filament fabrication slice the input 3D model into 2D layers and construct a toolpath. In prior work (ICRA 2016) we gave a simple greedy algorithm that changed this paradigm and constructed the toolpath in 3D by printing local features in their entirety. This algorithm significantly improved upon layer-based methods, achieving a 34% mean reduction of wasted motion. In this paper we give a new algorithm that is more robust and achieves significantly better performance than the greedy approach. Our algorithm utilizes local search and nearly doubles our prior improvement, achieving a mean/median reduction of 62% over layer-based methods on the same benchmark of over 400 models. We also study toolpath optimality using a novel integer linear programming formulation. This formulation allows us to solve a linear programming relaxation that, while computationally intensive, can give us a lower bound on the optimal solution quality, giving us the ability to rigorously characterize solution quality for a given input model.

In this paper, we present an efficient Dynamic Programing framework for optimal planning and control of legged robots. First we formulate this problem as an optimal control problem for switched systems. Then we propose a multi-level optimization approach to find the optimal switching times and the optimal continuous control inputs. Through this scheme, the decomposed optimization can potentially be done more efficiently than the combined approach. Finally, we present a continuous-time constrained LQR algorithm which simultaneously optimizes the feedforward and feedback controller with O(n) time-complexity. In order to validate our approach, we show the performance of our framework on a quadrupedal robot. We choose the Center of Mass dynamics and the full kinematic formulation as the switched system model where the switching times as well as the contact forces and the joint velocities are optimized for different locomotion tasks such as gap crossing, walking and trotting.

We propose an intelligent visuomotor system that interacts with the environment and memorizes the consequences of actions. As more memories are recorded and more interactions are observed, the agent becomes more capable of predicting the consequences of actions and is, thus, better at planning sequences of actions to solve tasks. In previous work, we introduced the aspect transition graph (ATG) which represents how actions lead from one observation to another using a directed multi-graph. In this work, we propose a novel aspect representation based on hierarchical CNN features, learned with convolutional neural networks, that supports manipulation and captures the essential affordances of an object based on RGB-D images. In a traditional planning system, robots are given a pre-defined set of actions that take the robot from one symbolic state to another. However symbolic states often lack the flexibility to generalize across similar situations. Our proposed representation is grounded in the robot's observations and lies in a continuous space that allows the robot to handle similar unseen situations. The hierarchical CNN features within a representation also allow the robot to act precisely with respect to the spatial location of individual features. We evaluate the robustness of this representation using the Washington RGB-D Objects Dataset and show that it achieves state of the art results for instance pose estimation. We then test this representation in conjunction with an ATG on a drill grasping task on Robonaut-2. We show that given grasp, drag, and turn demonstrations on the drill, the robot is capable of planning sequences of learned actions to compensate for reachability constraints.

The lateral line sensory system found in fish and other aquatic organisms is believed to aid in various behavioral activities such as station keeping and wall detection. This paper presents results for an artificial lateral line designed for use in an underwater vehicle. Algorithms for leveraging this system to estimate hydrodynamic forces and for wall detection are presented and experimentally validated. Three sensor setups were built, two with fixed sensor positions and one with a novel, modular sensor block approach. The modular realization allows for flexibility in testing and can be integrated into the control system of the autonomous underwater vehicle built by our group. These lateral line sensory systems were experimentally validated and the results are presented.

Onboard vision sensing is a current challenge in micro-scale robotics. Small flapping-wing robots such as the RoboBee present significant constraints on power, weight, and image quality for an onboard vision sensor. Here we report the integration of a 1 × 1 × 1.7 mm camera capable of video capture in flight. Inspired by gaze stabilization in insects, we designed and fabricated a one degree of freedom mechanism attached to the top of the RoboBee that achieves output angles of -41° to +60°. We perform open-loop roll maneuvers and demonstrate initial control of the gaze angle during flight. This represents the first use of a camera in free flight at this scale.

We study the problem of searching and tracking a collection of moving targets using a robot with a limited Field-of-View (FoV) sensor. The actual number of targets present in the environment is not known a priori. We propose a search and tracking framework based on the concept of Bayesian Random Finite Sets (RFSs). Specifically, we generalize the Gaussian Mixture Probability Hypothesis Density (GM-PHD) filter which was previously applied for only tracking problems to allow for simultaneous search and tracking. The proposed framework can extract individual target tracks as well as estimate the number and spatial density of the targets. We also show how to use Gaussian Process (GP) regression to extract and predict nonlinear target trajectories in this framework. We demonstrate the efficacy of our techniques through representative simulations where we also compare the performance of two active control strategies.

In this paper, we design provably-good algorithms for task allocation in multi-robot systems in the presence of payoff uncertainty. We consider a group of robots that has to perform a given set of tasks where each robot performs at most one task. The payoffs of the robots doing the tasks are assumed to be Gaussian random variables with known mean and variances. The total payoff of the robots is a sum of the individual payoffs of all the robots. The goal is to find an assignment with maximum payoff that can be achieved with a specified probability irrespective of the realization of the random variable. This problem can be formulated as a chance constrained combinatorial optimization problem. We develop a novel deterministic technique to solve this chance constrained optimization problem that ensures that the chance constraints are always satisfied. Adopting the notion of risk-aversion from the economics literature, we formulate a risk-averse task allocation problem, which is a deterministic integer optimization problem. We prove that by repeatedly solving the risk-averse task allocation problem using a one-dimensional search on the risk aversion parameter we find a solution for the chance constrained optimization formulation of the linear assignment problem with uncertain payoffs. We provide simulation results on randomly generated data to demonstrate our approach and also compare our method to existing approaches.

We address one of the main challenges towards autonomous quadrotor flight in complex environments, which is flight through narrow gaps. While previous works relied on off-board localization systems or on accurate prior knowledge of the gap position and orientation in the world reference frame, we rely solely on onboard sensing and computing and estimate the full state by fusing gap detection from a single onboard camera with an IMU. This problem is challenging for two reasons: (i) the quadrotor pose uncertainty with respect to the gap increases quadratically with the distance from the gap; (ii) the quadrotor has to actively control its orientation towards the gap to enable state estimation (i.e., active vision). We solve this problem by generating a trajectory that considers geometric, dynamic, and perception constraints: during the approach maneuver, the quadrotor always faces the gap to allow state estimation, while respecting the vehicle dynamics; during the traverse through the gap, the distance of the quadrotor to the edges of the gap is maximized. Furthermore, we replan the trajectory during its execution to cope with the varying uncertainty of the state estimate. We successfully evaluate and demonstrate the proposed approach in many real experiments, achieving a success rate of 80% and gap orientations up to 45°. To the best of our knowledge, this is the first work that addresses and achieves autonomous, aggressive flight through narrow gaps using only onboard sensing and computing and without prior knowledge of the pose of the gap.

Autonomous robot navigation through unknown, cluttered environments at high-speeds is still an open problem. Quadrotor platforms with this capability have only begun to emerge with the advancements in light-weight, small form factor sensing and computing. Many of the existing platforms, however, require excessive computation time to perform collision avoidance, which ultimately limits the vehicle's top speed. This work presents an efficient perception and planning approach that significantly reduces the computation time by using instantaneous perception data for collision avoidance. Minimum-time, state and input constrained motion primitives are generated by sampling terminal states until a collision-free path is found. The worst case performance of the Triple Integrator Planner (TIP) is nearly an order of magnitude faster than the state-of-the-art. Experimental results demonstrate the algorithm's ability to plan and execute aggressive collision avoidance maneuvers in highly cluttered environments.

Autonomous delivery of goods using a Micro Air Vehicle (MAV) is a difficult problem, as it poses high demand on the MAV's control, perception and manipulation capabilities. This problem is especially challenging if the exact shape, location and configuration of the objects are unknown. In this paper, we report our findings during the development and evaluation of a fully integrated system that is energy efficient and enables MAVs to pick up and deliver objects with partly ferrous surface of varying shapes and weights. This is achieved by using a novel combination of an electro-permanent magnetic gripper with a passively compliant structure and integration with detection, control and servo positioning algorithms. The system's ability to grasp stationary and moving objects was tested, as well as its ability to cope with different shapes of the object and external disturbances. We show that such a system can be successfully deployed in scenarios where an object with partly ferrous parts needs to be gripped and placed in a predetermined location.

This paper concentrates on design of a vision-based guidance command for aerial manipulation of a cylindrical object, using a stochastic model predictive approach. We first develop an image-based cylinder detection algorithm that utilizes a geometric characteristic of perspectively projected circles in 3D space. To enforce the object to be located inside sight of a camera, we formulate a visual servoing problem as a stochastic model predictive control (MPC) framework. By regarding x and y axes rotational velocities as stochastic variables, we guarantee the visibility of the camera considering underactuation of the system. We also provide experimental results that validate effectiveness of the proposed algorithm.

In physical human-robot interaction, the coexistence of robots and humans in the same workspace requires the guarantee of a stable interaction, trying to minimize the effort for the operator. To this aim, the admittance control is widely used and the appropriate selection of the its parameters is crucial, since they affect both the stability and the ability of the robot to interact with the user. In this paper, we present a strategy for detecting deviations from the nominal behavior of an admittance-controlled robot and for adapting the parameters of the controller while guaranteeing the passivity. The proposed methodology is validated on a KUKA LWR 4+.

In this paper, the problems associated with accurate path tracking control in off-road conditions is addressed with model-based adaptive control. In particular, the estimation of grip conditions is investigated through the derivation of a new observer and by gathering kinematic and dynamic models into a single framework. This new reference point employs a unique observer regardless of the velocity of the robots. Previous approaches necessitated the switching of models depending upon the phenomena encountered as well as robot dynamics. The observer proposed here allows an accurate and reactive estimation of sliding. This permits to feed relevantly a control law based on an extended kinematic model, enabling accurate path tracking, even in harsh conditions and when facing significant dynamic effects such as spin around.

This paper presents a novel adaptive approach to fast sampling-based motion planning by learning models of collision and collision-free regions in configuration spaces in an online manner. The proposed approach incrementally learns Gaussian Mixture Models (GMMs) for collision detection in high dimensional configuration spaces. In practical applications for robotic manipulation, the representation of collision and collision-free regions in configuration space can change due to relative motion between the robot base and workspace. We show how to rapidly adapt to such changes by using inverse kinematics to transform the parameters of the Gaussian mixture model to new configurations. The transformed model is initially used as a prior and then continually updated and refined as the RRT planning algorithm proceeds in real-time. This approach is extremely computationally efficient, and our proposed method is compared with traditional sampling-based planning methods on a number of experimental robot arm planning scenarios.

This paper introduces the adaptive bias and adaptive gain (ABAG) algorithm for closed-loop electronic speed control (ESC) of the brushless direct current (BLDC) motors typically used to spin the propellers in multi-rotor aerial robots. The ABAG algorithm is adaptive and robust in the sense that it does not require the knowledge of any mechanical/electrical parameter of the motor/propeller group and that neither a pre-calibration nor the knowledge of the feedforward/nominal input is needed. The ABAG algorithm is amenable to an extremely low complexity implementation. We experimentally prove that it can run in 27.5 μs on a 8 MHz microcontroller with no floating point unit and limited arithmetic capabilities allowing only 8-bit additions, subtractions and multiplications. Besides the controller implementation we present a self-contained open source software architecture that handles the entire speed control process, including clock synchronization, and over-current and blockage safeties. The excellent performance and robustness of ABAG are shown by experimental tests and aerial physical interaction experiments.

Odor source localization with mobile robots has recently been subject to many research works, but remains a challenging task mainly due to the large number of environmental parameters that make it hard to describe gas concentration fields. We designed a new algorithm called Adaptive Lévy Taxis (ALT) to achieve odor plume tracking through a correlated random walk. In order to compare its performances with well-established solutions, we have implemented three moth-inspired algorithms on the same robotic platform. To improve the performance of the latter algorithms, we developed a rigorous way to determine one of their key parameters, the odor concentration threshold at which the robot considers to be inside or outside the plume. The methods have been systematically evaluated in a large wind tunnel under various environmental conditions. Experiments revealed that the performance of ALT is consistently good in all environmental conditions (in particular when compared to the three reference algorithms) in terms of both distance traveled to find the source and success rate.

We present an approach to learning control policies for physical robots that achieves high efficiency by adjusting existing policies that have been learned on similar source systems, such as a similar robot with different physical parameters, or an approximate dynamics model simulator. This can be viewed as calibrating a policy learned on a source system, to match a desired behaviour in similar target systems. Our approach assumes that the trajectories described by the source robot are feasible on the target robot. By making this assumption, we only need to learn a mapping from the source robot state and action spaces to the target robot action space, which we call a policy adjustment model. We demonstrate our approach in simulation in the cart-pole balancing task and a two link double pendulum. We also validate our approach with a physical cart-pole system, where we adjust a learned policy under changes to the weight of the pole.