Jason Pusey

Towards a Comparative Measure of Legged Agility

We introduce an agility measure enabling the comparison of two very different leaping-from-rest transitions by two comparably powered but morphologically different legged robots. We use the measure to show that a flexible spine outperforms a rigid back in the leaping-from-rest task. The agility measure also sheds light on the source of this benefit: core actuation through a sufficiently powerful parallel elastic actuated spine outperforms a similar power budget applied either only to preload the spine or only to actuate the spine during the leap, as well as a rigid backed configuration of the identical machine.

Free-Standing Leaping Experiments with a Power-Autonomous, Elastic-Spined Quadruped

We document initial experiments with Canid, a freestanding, power-autonomous quadrupedal robot equipped with a parallel actuated elastic spine. Research into robotic bounding and galloping platforms holds scientific and engineering interest because it can both probe biological hypotheses regarding bounding and galloping mammals and also provide the engineering community with a new class of agile, efficient and rapidly-locomoting legged robots. We detail the design features of Canid that promote our goals of agile operation in a relatively cheap, conventionally prototyped, commercial off-the-shelf actuated platform. We introduce new measurement methodology aimed at capturing our robot’s “body energy” during real time operation as a means of quantifying its potential for agile behavior. Finally, we present joint motor, inertial and motion capture data taken from Canid’s initial leaps into highly energetic regimes exhibiting large accelerations that illustrate the use of this measure and suggest its future potential as a platform for developing efficient, stable, hence useful bounding gaits.

Generalized predictive control algorithm of a simplified ground vehicle suspension system

This paper discusses research conducted by the Army Research Laboratory (ARL) - Vehicle Technology Directorate (VTD) on advanced suspension control. ARL-VTD has conducted research on advanced suspension systems that will reduce the chassis vibration of ground vehicles while maintaining tire contact with the road surface. The purpose of this research is to reduce vibration-induced fatigue to the Warfighter as well as to improve the target aiming precision in-theater. The objective of this paper was to explore the performance effectiveness of various formulations of the generalized predictive control algorithm in a simulation environment. Each version of the control algorithm was applied to an identical model subjected to the same ground disturbance input and compared to a baseline passive suspension system. The control algorithms considered include a generalized predictive controller (GPC) with implicit disturbances, GPC with explicit disturbances, and GPC with preview control. The suspension model used was a two-degree-of-freedom dof quarter-car model with a given set of vehicle parameters. The performance of the control algorithms were compared based on their effectiveness in controlling peak acceleration and overall average acceleration over a range of vehicle speeds. The algorithms demonstrated significant reductions in the chassis acceleration of the quarter-car model.

Tools for the design of stable yet nonsteady bounding control

Control efforts on dynamic gaits for legged systems traditionally focus on limit cycles and their stability. However, there are many practical situations where step-to-step variability is highly desirable, for example to achieve variable footholds or to recover and replan after perturbations. In this paper we present an effective, high-level switching control framework for overcoming terrain obstacles using the familiar A* algorithm to search a mesh over the reachable space for a given set of controllers. In support of this, we present new low-level control strategies for generating stable bounding with planar models of a spring-legged quadruped robot, and demonstrate their use crossing gaps in the terrain.

Validation and verification of a high-fidelity computational model for a bounding robot's parallel actuated elastic spine

We document the design and preliminary numerical simulation study of a high fidelity model of Canid, a recently introduced bounding robot. Canid is a free-standing, power-autonomous quadrupedal machine constructed from standard commercially available electromechanical and structural elements, incorporating compliant C-shaped legs like those of the decade old RHex design, but departing from that standard (and, to the best of our knowledge, from any prior) robot platform in its parallel actuated elastic spine. We have used a commercial modeling package to develop a finite-element model of the actuated, cable-driven, rigid-plate-reinforced harness for the carbon-fiber spring that joins the robot’s fore- and hind-quarters. We compare a numerical model of this parallel actuated elastic spine with empirical data from preliminary physical experiments with the most important component of the spine assembly: the composite leaf spring. Specifically, we report our progress in tuning the mechanical properties of a standard modal approximation to a conventional compliant beam model whose boundary conditions represent constraints imposed by the actuated cable driven vertebral plates that comprise the active control affordance over the spine. We conclude with a brief look ahead at near-term future experiments that will compare predictions of this fitted composite spring model with data taken from the physical spine flexed in isolation from the actuated harness.

Mesh-based switching control for robust and agile dynamic gaits

In this paper, we describe and analyze mesh-based tools to control bounding motions of an 8 degree-of-freedom planar quadruped model with limited footholds on terrain. There are two complementary goals in our presentation. First, we aim to clarify potential advantages and disadvantages of our mesh-based approach in planning agile motions for a legged system. A key advantage is the ability to map the reachable states and their feasible transitions, given a relatively high-dimensional nonlinear dynamic system for which traditional meshing techniques would be impractical. A suspected disadvantage is that meshing has finite resolution, and robustness of mesh-based results should correspondingly be considered. Our second goal is to discuss appropriate frameworks for optimizing agility. Unlike typical locomotion optimization studies, in which control is designed for a limit cycle behavior that minimizes energy use or improves robustness to perturbations, here we focus on quantifying the performance of sets of controllers that together enhance reachability of the controlled system. In planning agile motions for our legged system model, we find that our mesh-based policies predict future dynamics robustly for plans up to about a 5-step horizon, and in quantifying controller sets, we emphasize that both the number of and parameterizations for such controllers should be considered in tandem during optimization.

A comparative study of leg kinematics for energy-efficient locomotion

It has been theorized that biological legs or the serial leg with a knee and hip joint has evolved over centuries for energy efficient locomotion and as such, has been adapted into a multitude of legged robots. However, recent success of legged robots with alternate leg morphologies without actuated knees, such as the parallel and symmetric five-bar link leg raises the question: which leg geometry is more energy-efficient and why? To answer this question, we created a minimal model of bipedal walking whose non-dimensionalised equations of motion have a single free parameter, the leg ratio and defined as the ratio of the distal to the proximal leg length. Then we performed an energy minimization for a given leg ratio and combination of speed and step length. When we optimized mechanical work, we found that all three legs have an identical efficiency, but the symmetric leg has the lowest peak torque. When we optimized a cost representative of an electric motor, we found that the serial leg is most energy efficient for all leg ratios, and the cost decreases as the leg ratio increases. For a leg ratio of 1, the parallel and symmetric leg have identical efficiencies. As the leg ratio increases, the efficiency of the symmetric leg approaches that of the serial leg while that of parallel leg decreases. However, the symmetric leg produces the least peak torque followed by the serial leg for leg ratios greater than 1. Our conclusion is that the symmetric leg rivals the serial leg by being easier to design and having smaller peak torques leading to smaller actuators at the cost of being slightly less energy-efficient.

Simulation tools for robotics research and assessment

The Robotics Collaborative Technology Alliance (RCTA) program focuses on four overlapping technology areas: Perception, Intelligence, Human-Robot Interaction (HRI), and Dexterous Manipulation and Unique Mobility (DMUM). In addition, the RCTA program has a requirement to assess progress of this research in standalone as well as integrated form. Since the research is evolving and the robotic platforms with unique mobility and dexterous manipulation are in the early development stage and very expensive, an alternate approach is needed for efficient assessment. Simulation of robotic systems, platforms, sensors, and algorithms, is an attractive alternative to expensive field-based testing. Simulation can provide insight during development and debugging unavailable by many other means. This paper explores the maturity of robotic simulation systems for applications to real-world problems in robotic systems research. Open source (such as Gazebo and Moby), commercial (Simulink, Actin, LMS), government (ANVEL/VANE), and the RCTA-developed RIVET simulation environments are examined with respect to their application in the robotic research domains of Perception, Intelligence, HRI, and DMUM. Tradeoffs for applications to representative problems from each domain are presented, along with known deficiencies and disadvantages. In particular, no single robotic simulation environment adequately covers the needs of the robotic researcher in all of the domains. Simulation for DMUM poses unique constraints on the development of physics-based computational models of the robot, the environment and objects within the environment, and the interactions between them. Most current robot simulations focus on quasi-static systems, but dynamic robotic motion places an increased emphasis on the accuracy of the computational models. In order to understand the interaction of dynamic multi-body systems, such as limbed robots, with the environment, it may be necessary to build component-level computational models to provide the necessary simulation fidelity for accuracy. However, the Perception domain remains the most problematic for adequate simulation performance due to the often cartoon nature of computer rendering and the inability to model realistic electromagnetic radiation effects, such as multiple reflections, in real-time.