There are many scenarios in which a mobile agent may not want its path to be predictable. Examples include preserving privacy or confusing an adversary. However, this desire for deception can conflict with the need for a low path cost. Optimal plans such as those produced by RRT∗ may have low path cost, but their optimality makes them predictable. Similarly, a deceptive path that features numerous zig-zags may take too long to reach the goal. We address this trade-off by drawing inspiration from adversarial machine learning. We propose a new planning algorithm, which we title Adversarial RRT*. Adversarial RRT∗ attempts to deceive machine learning classifiers by incorporating a predicted measure of deception into the planner cost function. Adversarial RRT∗ considers both path cost and a measure of predicted deceptiveness in order to produce a trajectory with low path cost that still has deceptive properties. We demonstrate the performance of Adversarial RRT*, with two measures of deception, using a simulated Dubins vehicle. We show how Adversarial RRT∗ can decrease cumulative RNN accuracy across paths to 10%, compared to 46% cumulative accuracy on near-optimal RRT∗ paths, while keeping path length within 16% of optimal. We also present an example demonstration where the Adversarial RRT∗ planner attempts to safely deliver a high value package while an adversary observes the path and tries to intercept the package.
This paper describes how the performance of motion primitive-based planning algorithms can be improved using reinforcement learning. Specifically, we describe and evaluate a framework that autonomously improves the performance of a primitive-based motion planner. The improvement process consists of three phases: exploration, extraction, and reward updates. This process can be iterated continuously to provide successive improvement. The exploration step generates new trajectories, and the extraction step identifies new primitives from these trajectories. These primitives are then used to update rewards for continued exploration. This framework required novel shaping rewards, development of a primitive extraction algorithm, and modification of the Hybrid A* algorithm. The framework is tested on a navigation task using a nonlinear F-16 model. The framework autonomously added 91 motion primitives to the primitive library and reduced average path cost by 21.6 s, or 35.75% of the original cost. The learned primitives are applied to an obstacle field navigation task, which was not used in training, and reduced path cost by 16.3 s, or 24.1%. Additionally, two heuristics for the modified Hybrid A* algorithm are designed to improve effective branching factor.
Wellbore integrity is a significant problem in the U.S. and worldwide, which has serious adverse environmental and energy security consequences. Wells are constructed with a cement barrier designed to last about 50 years. Indirect measurements and models are commonly used to identify wellbore damage and leakage, often producing subjective and even erroneous results. The research presented herein focuses on new technologies to improve monitoring and detection of wellbore failures (leaks) by developing a multi-step machine learning approach to localize two types of thermal defects within a wellbore model, a prototype mechatronic system for automatically drilling small diameter holes of arbitrary depth to monitor the integrity of oil and gas wells in situ, and benchtop testing and analyses to support the development of an autonomous real-time diagnostic tool to enable sensor emplacement for monitoring wellbore integrity. Each technology was supported by experimental results. This research has provided tools to aid in the detection of wellbore leaks and significantly enhanced our understanding of the interaction between small-hole drilling and wellbore materials.
One of the greatest barriers to geothermal energy expansion is the high cost of drilling during exploration, assessment, and monitoring. Microhole drilling technology—small-diameter 2–4 in. (~5.1–10.2 cm) boreholes—is one potential low-cost alternative for monitoring and evaluating bores. However, delivering high weight-on-bit (WOB), high torque rotational horsepower to a conventional drill bit does not scale down to the hole sizes needed to realize the cost savings. Coiled tube drilling technology is one solution, but these systems are limited by the torque resistance of the coil system, helical buckling in compression, and most of all, WOB management. The evaluation presented herein will: (i) evaluate the technical and economic feasibility of low WOB technologies (specifically, a percussive hammer and a laser-mechanical system), (ii) develop downhole rotational solutions for low WOB drilling, (iii) provide specifications for a low WOB microhole drilling system, (iv) implement WOB control for low WOB drilling, and (v) evaluate and test low WOB drilling technologies.
The well documented promise of microholes has not yet matched expectations. A fundamental issue is that delivering high weight-on-bit (WOB), high torque rotational horsepower to a conventional drill bit does not scale down to the hole sizes necessary to realize the envisioned cost savings. Prior work has focused on miniaturizing the various systems used in conventional drilling technologies, such as motors, steering systems, mud handling and logging tools, and coiled tubing drilling units. As smaller diameters are targeted for these low WOB drilling technologies, several associated sets of challenges arise. For example, energy transfer efficiency in small diameter percussive hammers is different than conventional hammers. Finding adequate methods of generating rotation at the bit are also more difficult. A low weight-on-bit microhole drilling system was proposed, conceived, and tested on a limited scale. The utility of a microhole was quantified using flow analyses to establish bounds for usable microholes. Two low weight-on-bit rock reduction techniques were evaluated and developed, including a low technology readiness level concept in the laser-assisted mechanical drill and a modified commercial percussive hammer. Supporting equipment, including downhole rotation and a drill string twist reaction tool, were developed to enable wireline deployment of a drilling assembly. Although the various subsystems were tested and shown to work well individually in a laboratory environment, there is still room for improvement before the microhole drilling system is ready to be deployed. Ruggedizing the various components will be key, as well as having additional capacity in a conveyance system to provide additional capacity for pullback and deployment.
Legged humanoid robots promise revolutionary mobility and effectiveness in environments built for humans. However, inefficient use of energy significantly limits their practical adoption. The humanoid biped walking anthropomorphic novelly-driven efficient robot for emergency response (WANDERER) achieves versatile, efficient mobility, and high endurance via novel drive-trains and passive joint mechanisms. Results of a test in which WANDERER walked for more than 4 h and covered 2.8 km on a treadmill, are presented. Results of laboratory experiments showing even more efficient walking are also presented and analyzed in this article. WANDERER's energetic performance and endurance are believed to exceed the prior literature in human-scale humanoid robots. This article describes WANDERER, the analytical methods and innovations that enable its design, and system-level energy efficiency results.
Legged robots promise radical mobility for challenging environments, but must be made more energy efficient to be practical. Historically, legged robot design has required efficiency to be traded against versatility. Much energy is lost in actuators and transmissions because few actuation systems are capable of operating efficiently across the wide range of operating conditions (e.g. different joint speeds and torques) required for legged locomotion. We describe a drivetrain topology that overcomes many of these limitations. Our approach combines high-torque electromagnetic motors and low-loss transmissions with a tailored and adjustable set of joint-specific passive mechanisms called support elements, which modulate the energy flow between motors and joints to minimize the electrical energy consumed. We present an optimization-based design method that draws on available bipedal gait data to select optimal support element configurations and parameters. Simple adjustments may be made to support elements at certain joints to enable a wide variety of locomotion with high efficiency. We present results, specific to the 3D humanoid bipedal STEPPR robot, in which support elements are co-optimized across a library of several gaits, converging on a set of designs that predict an average reduction of electrical energy of more than 50% across a set of 15 gaits, with energy savings reaching as much as 85% for some gaits. Concepts were prototyped and tested on a bench testbed, validating the predicted energy savings. Support elements were implemented on STEPPR, and energy savings of more than 35% were demonstrated.
The ability to rapidly drill through diverse, layered materials can greatly enhance future mine-rescue operations, energy exploration, and underground operations. Pneumatic-percussive drilling holds great promise in this area due to its ability to penetrate very hard materials and potential for portability. Currently such systems require expert operators who require extensive training. We envision future applications where first responders who lack such training can still respond rapidly and safely perform operations. Automated techniques can reduce the dependence on expert operators while increasing efficiency and safety. However, current progress in this area is restricted by the difficulty controlling such systems and the complexity of modeling percussive rock-bit interactions. In this work we develop and experimentally validate a novel intelligent percussive drilling architecture that is tailored to autonomously operate in diverse, layered materials. Our approach combines low-level feedback control, machine learning-based material classification, and on-line optimization. Our experimental results demonstrate the effectiveness of this approach and illustrate the performance benefits over conventional methods.
Epoxies and resins can require careful temperature sensing and control in order to monitor and prevent degradation. To sense the temperature inside a mold, it is desirable to utilize a small, wireless sensing element. In this paper, we describe a new architecture for wireless temperature sensing and closed-loop temperature control of exothermic polymers. This architecture is the first to utilize magnetic field estimates of the temperature of permanent magnets within a temperature feedback control loop. We further improve performance and applicability by demonstrating sensing performance at relevant temperatures, incorporating a cure estimator, and implementing a nonlinear temperature controller. This novel architecture enables unique experimental results featuring closed-loop control of an exothermic resin without any physical connection to the inside of the mold. In this paper, we describe each of the unique features of this approach, including magnetic field-based temperature sensing, extended Kalman filtering for cure state estimation, and nonlinear feedback control over time-varying temperature trajectories. We use experimental results to demonstrate how low-cost permanent magnets can provide wireless temperature sensing up to ∼ 90°C. In addition, we use a polymer cure-control testbed to illustrate how internal temperature sensing can provide improved temperature control over both short and long time-scales. This wireless temperature sensing and control architecture holds value for a range of manufacturing applications.
Vibration sensing is critical for a variety of applications from structural fatigue monitoring to understanding the modes of airplane wings. In particular, remote sensing techniques are needed for measuring the vibrations of multiple points simultaneously, assessing vibrations inside opaque metal vessels, and sensing through smoke clouds and other optically challenging environments. In this paper, we propose a method which measures high-frequency displacements remotely using changes in the magnetic field generated by permanent magnets. We leverage the unique nature of vibration tracking and use a calibrated local model technique developed specifically to improve the frequency-domain estimation accuracy. The results show that two-dimensional local models surpass the dipole model in tracking high-frequency motions. A theoretical basis for understanding the effects of electronic noise and error due to correlated variables is generated in order to predict the performance of experiments prior to implementation. Simultaneous measurements of up to three independent vibrating components are shown. The relative accuracy of the magnet-based displacement tracking with respect to the video tracking ranges from 40 to 190 μ m when the maximum displacements approach ±5 mm and when sensor-to-magnet distances vary from 25 to 36 mm. Last, vibration sensing inside an opaque metal vessel and mode shape changes due to damage on an aluminum beam are also studied using the wireless permanent-magnet vibration sensing scheme.
Torque feedback control and series elastic actuators are widely used to enable compact, highly-geared electric motors to provide low and controllable mechanical impedance. While these approaches provide certain benefits for control, their impact on system energy consumption is not widely understood. This paper presents a model for examining the energy consumption of drivetrains implementing various target dynamic behaviors in the presence of gear reductions and torque feedback. Analysis of this model reveals that under cyclical motions for many conditions, increasing the gear ratio results in greater energy loss. A similar model is presented for series elastic actuators and used to determine the energy consequences of various spring stiffness values. Both models enable the computation and optimization of power based on specific hardware manifestations, and illustrate how energy consumption sometimes defies conventional best-practices. Results of evaluating these two topologies as part of a drivetrain design optimization for two energy-efficient electrically driven humanoids are summarized. The model presented enables robot designers to predict the energy consequences of gearing and series elasticity for future robot designs, helping to avoid substantial energy sinks that may be inadvertently introduced if these issues are not properly analyzed.
Drilling is a repetitive, dangerous and costly process and a strong candidate for automation. We describe a method for autonomously controlling a rotary drilling process as it transitions through multiple materials with very different dynamics. This approach classifies the drilling medium based on real-time measurements and comparison to prior drilling data, and can identify the material type, drilling region, and approximately optimal set-point based on data from as few as one operating condition. The controller uses these set-points as initial conditions, and then conducts an optimal search to maximize performance, e.g. by minimizing mechanical specific energy. The control architecture is described, and the material estimation process is detailed. The results of experiments that implement autonomous drilling through a layered concrete and granite sample are discussed.
This paper describes the design and performance of a synthetic rope on sheave drive system. This system uses synthetic ropes instead of steel cables to achieve low weight and a compact form factor. We demonstrate how this system is capable of 28-Hz torque control bandwidth, 95% efficiency, and quiet operation, making it ideal for use on legged robots and other dynamic physically interactive systems. Component geometry and tailored maintenance procedures are used to achieve high endurance. Endurance tests based on walking data predict that the ropes will survive roughly 247,000 cycles when used on large (90 kg), fully actuated bipedal robot systems. The drive systems have been incorporated into two novel bipedal robots capable of three-dimensional unsupported walking. Robot data illustrate effective torque tracking and nearly silent operation. Finally, comparisons with alternative transmission designs illustrate the size, weight, and endurance advantages of using this type of synthetic rope drive system.
This paper describes how parallel elastic elements can be used to reduce energy consumption in the electric-motor-driven, fully actuated, Sandia Transmission-Efficient Prototype Promoting Research (STEPPR) bipedal walking robot without compromising or significantly limiting locomotive behaviors. A physically motivated approach is used to illustrate how selectively engaging springs for hip adduction and ankle flexion predict benefits for three different flat-ground walking gaits: human walking, human-like robot walking, and crouched robot walking. Based on locomotion data, springs are designed and substantial reductions in power consumption are demonstrated using a bench dynamometer. These lessons are then applied to STEPPR, a fully actuated bipedal robot designed to explore the impact of tailored joint mechanisms on walking efficiency. Featuring high-Torque brushless DC motors, efficient low-ratio transmissions, and high-fidelity torque control, STEPPR provides the ability to incorporate novel joint-level mechanisms without dramatically altering high-level control. Unique parallel elastic designs are incorporated into STEPPR, and walking data show that hip adduction and ankle flexion springs significantly reduce the required actuator energy at those joints for several gaits. These results suggest that parallel joint springs offer a promising means of supporting quasi-static joint torques due to body mass during walking, relieving motors of the need to support these torques and substantially improving locomotive energy efficiency.