Nicolas Hudson

The next best touch for model-based localization

This paper introduces a tactile or contact method whereby an autonomous robot equipped with suitable sensors can choose the next sensing action involving touch in order to accurately localize an object in its environment. The method uses an information gain metric based on the uncertainty of the objects pose to determine the next best touching action. Intuitively, the optimal action is the one that is the most informative. The action is then carried out and the state of the objects pose is updated using an estimator. The method is further extended to choose the most informative action to simultaneously localize and estimate the objects model parameter or model class. Results are presented both in simulation and in experiment on the DARPA Autonomous Robotic Manipulation Software (ARM-S) robot.

Feature and pose constrained visual Aided Inertial Navigation for computationally constrained aerial vehicles

A Feature and Pose Constrained Extended Kalman Filter (FPC-EKF) is developed for highly dynamic computationally constrained micro aerial vehicles. Vehicle localization is achieved using only a low performance inertial measurement unit and a single camera. The FPC-EKF framework augments the vehicles state with both previous vehicle poses and critical environmental features, including vertical edges. This filter framework efficiently incorporates measurements from hundreds of opportunistic visual features to constrain the motion estimate, while allowing navigating and sustained tracking with respect to a few persistent features. In addition, vertical features in the environment are opportunistically used to provide global attitude references. Accurate pose estimation is demonstrated on a sequence including fast traversing, where visual features enter and exit the field-of-view quickly, as well as hover and ingress maneuvers where drift free navigation is achieved with respect to the environment.

Experimental results of rover-based coring and caching

Experimental results are presented for experiments performed using a prototype rover-based sample coring and caching system. The system consists of a rotary percussive coring tool on a five degree-of-freedom manipulator arm mounted on a FIDO-class rover and a sample caching subsystem mounted on the rover. Coring and caching experiments were performed in a laboratory setting and in a field test at Mono Lake, California. Rock abrasion experiments using an abrading bit on the coring tool were also performed. The experiments indicate that the sample acquisition and caching architecture is viable for use in a 2018 timeframe Mars caching mission and that rock abrasion using an abrading bit may be feasible in place of a dedicated rock abrasion tool.1 2

Estimation and control for autonomous coring from a rover manipulator

A system consisting of a set of estimators and autonomous behaviors has been developed which allows robust coring from a low-mass rover platform, while accommodating for moderate rover slip. A redundant set of sensors, including a force-torque sensor, visual odometry, and accelerometers are used to monitor discrete critical and operational modes, as well as to estimate continuous drill parameters during the coring process. A set of critical failure modes pertinent to shallow coring from a mobile platform is defined, and autonomous behaviors associated with each critical mode are used to maintain nominal coring conditions. Autonomous shallow coring is demonstrated from a low-mass rover using a rotary-percussive coring tool mounted on a 5 degree-of-freedom (DOF) arm. A new architecture of using an arm-stabilized, rotary percussive tool with the robotic arm used to provide the drill z-axis linear feed is validated. Particular attention to hole start using this architecture is addressed. An end-to-end coring sequence is demonstrated, where the rover autonomously detects and then recovers from a series of slip events that exceeded 9 cm total displacement.

A Quadratic Programming Approach to Quasi-Static Whole-Body Manipulation

This paper introduces a local motion planning method for robotic systems with manipulating limbs, moving bases (legged or wheeled), and stance stability constraints arising from the presence of gravity. We formulate the problem of selecting local motions as a linearly constrained quadratic program (QP), that can be solved efficiently. The solution to this QP is a tuple of locally optimal joint velocities. By using these velocities to step towards a goal, both a path and an inverse-kinematic solution to the goal are obtained. This formulation can be used directly for real-time control, or as a local motion planner to connect waypoints. This method is particularly useful for high-degree-of-freedom mobile robotic systems, as the QP solution scales well with the number of joints. We also show how a number of practically important geometric constraints (collision avoidance, mechanism self-collision avoidance, gaze direction, etc.) can be readily incorporated into either the constraint or objective parts of the formulation. Additionally, motion of the base, a particular joint, or a particular link can be encouraged/discouraged as desired. We summarize the important kinematic variables of the formulation, including the stance Jacobian, the reach Jacobian, and a center of mass Jacobian. The method is easily extended to provide sparse solutions, where the fewest number of joints are moved, by iteration using Tibshirani’s method to accommodate an \(l_1\) regularizer. The approach is validated and demonstrated on SURROGATE, a mobile robot with a TALON base, a 7 DOF serial-revolute torso, and two 7 DOF modular arms developed at JPL/Caltech.

Dual arm estimation for coordinated bimanual manipulation

This paper develops an estimation framework for sensor-guided dual-arm manipulation of a rigid object. Using an unscented Kalman Filter (UKF), the approach combines both visual and kinesthetic information to track both the manipulators and object. From visual updates of the object and manipulators, and tactile updates, the method estimates both the robots internal state and the objects pose. Nonlinear constraints are incorporated into the framework to deal with the an additional arm and ensure the state is consistent. Two frameworks are compared in which the first framework run two single arm filters in parallel and the second consists of the augment dual arm filter with nonlinear constraints. Experiments on a wheel changing task are demonstrated using the DARPA ARM-S system, consisting of dual Barrett- WAM manipulators.

A stochastic framework for hybrid system identification with application to neurophysiological systems

This paper adapts the Gibbs sampling method to the problem of hybrid system identification. We define a Generalized Linear Hiddenl Markov Model (GLHMM) that combines switching dynamics from Hidden Markov Models, with a Generalized Linear Model (GLM) to govern the continuous dynamics. This class of models, which includes conventional ARX models as a special case, is particularly well suited to this identification approach. Our use of GLMs is also driven by potential applications of this approach to the field of neural prosthetics, where neural Poisson-GLMs can model neural firing behavior. The paper gives a concrete algorithm for identification, and an example motivated by neuroprosthetic considerations.

Learning Hybrid System Models for Supervisory Decoding of Discrete State, with applications to the Parietal Reach Region

Based on Gibbs sampling, a novel method to identify mathematical models of neural activity in response to temporal changes of behavioral or cognitive state is presented. This work is motivated by the developing field of neural prosthetics, where a supervisory controller is required to classify activity of a brain region into suitable discrete modes. Here, neural activity in each discrete mode is modeled with nonstationary point processes, and transitions between modes are modeled as hidden Markov models. The effectiveness of this framework is first demonstrated on a simulated example. The identification algorithm is then applied to extracellular neural activity recorded from multi-electrode arrays in the parietal reach region of a rhesus monkey, and the results demonstrate the ability to decode discrete changes even from small data sets

Rover reconfiguration for body-mounted coring with slip

Active and passive approaches to accommodating moderate rover slip during coring, using a body mounted coring tool were studied. The work addresses the possibility of a 200kg rover experiencing moderate slip such as 1cm/minute while coring on a Martian slope using a body-mounted coring tool in a possible future Mars Sample Return mission. Rover slip while coring could be accommodated with passive compliance or active rover reconfiguration. A passive compliance device was designed that constrains the compliant motion of the tool relative to the rover to a plane with travel of approximately 1.5cm. Active reconfiguration relies upon wheel motion and a transverse linear stage to provide actuation of the coring tool position in the nominal wheel plane. The rover actively reconfigures utilizing six axis force feedback of the forces and torques on the coring tool. Rover slip is measured using Absolute Motion Visual Odometry (AMVO). Rover reconfiguration is demonstrated while coring on a slope with slip. However, the implemented body mounted coring tool approach is fundamentally limited: First, no out-ofplane actuation is considered, negating the possibility of rover roll and pitch compensation. Second, wheel-ground interaction could cause unintended system responses when wheel motion occurs, including motion in the uncontrollable roll and pitch degrees-of-freedom.

Probabilistic round trip contamination analysis of a Mars Sample acquisition and handling process using Markovian decompositions

A method for evaluating the probability of a Viable Earth Microorganism (VEM) contaminating a sample during the sample acquisition and handling (SAH) process of a potential future Mars Sample Return mission is developed. A scenario where multiple core samples would be acquired using a rotary percussive coring tool, deployed from an arm on a MER class rover is analyzed. The analysis is conducted in a structured way by decomposing sample acquisition and handling process into a series of discrete time steps, and breaking the physical system into a set of relevant components. At each discrete time step, two key functions are defined: The probability of a VEM being released from each component, and the transport matrix, which represents the probability of VEM transport from one component to another. By defining the expected the number of VEMs on each component at the start of the sampling process, these decompositions allow the expected number of VEMs on each component at each sampling step to be represented as a Markov chain. This formalism provides a rigorous mathematical framework in which to analyze the probability of a VEM entering the sample chain, as well as making the analysis tractable by breaking the process down into small analyzable steps.