Go Fetch! - Dynamic Grasps using Boston Dynamics Spot with External Robotic Arm

Abstract

We combine Boston Dynamics Spot R with a lightweight, external robot arm to perform dynamic grasping maneuvers. While Spot is a reliable, robust and easy-to-control mobile robot, these highly desirable qualities come with the price that the control access granted to the user is restricted. Consequently Spot's behavior must largely be treated as a black box, which causes difficulties when combined with a moving payload such as a robotic arm. We overcome the arising challenges by building a model of the combined platform, fitting the corresponding model parameters using experimental data and a straight-forward optimization framework. We use this model to generate control commands for the physical platform using trajectory optimization. We demonstrate that even with a simple model, and control trajectories deployed in a feedforward manner, the combined platform is capable of executing grasping tasks in a dynamic fashion. Furthermore, we show how the platform can use the additional degrees of freedom of the legs to extend the reachability of the arm.

RoLoMa: Robust Loco-Manipulation for Quadruped Robots with Arms

“…A few years later, Zimmermann et al [2] equipped Boston Dynamics' flagship quadruped robot Spot with a Kinova arm in order to perform dynamic grasping maneuvers. Direct control of Spot's actuated joints is not possible because of restricted access to its low-level controller.…”

“…1), has four legs and one arm attached to its base; each leg has three motors and the arm has six. 2 We describe the motion of the system with respect to (w.r.t.) a fixed inertial frame I.…”

“…where M (q) ∈ R nv×nv is the mass matrix, and h(q, v) ∈ R nv is the vector of Coriolis, centrifugal, and gravity terms. On the right-hand side of the equation, τ ∈ R nτ is the vector of joint torques commanded to the system, and the selection 2 Please note that our formulation is not tied to the specific robot shown in Fig. 1.…”

Trajectory Optimization Using Neural Network Gradients of Learned Dynamics

“…MBRL has been successfully applied to a variety of robot control tasks in simulation and hardware [19]- [27]. In particular, [19]- [23] use gradient-based optimization schemes to obtain the control policy. Nonetheless, these methods either use parameterized models with hand-picked features [23] or Gaussian processes (GPs) [28] to learn the system dynamics.…”

“…In particular, [19]- [23] use gradient-based optimization schemes to obtain the control policy. Nonetheless, these methods either use parameterized models with hand-picked features [23] or Gaussian processes (GPs) [28] to learn the system dynamics. Hand-picking features is a time-consuming, often unintuitive, and sometimes also an impossible process.…”