Forestry machines are heavy vehicles performing complex manipulation tasks in unstructured production forest environments. Together with the complex dynamics of the onboard hydraulically actuated cranes, the rough forest terrains have posed a particular challenge in forestry automation. In this study, the feasibility of applying reinforcement learning control to forestry crane manipulators is investigated in a simulated environment. Our results show that it is possible to learn successful actuator-space control policies for energy efficient log grasping by invoking a simple curriculum in a deep reinforcement learning setup. Given the pose of the selected logs, our best control policy reaches a grasping success rate of 97%. Including an energy-optimization goal in the reward function, the energy consumption is significantly reduced compared to control policies learned without incentive for energy optimization, while the increase in cycle time is marginal. The energy-optimization effects can be observed in the overall smoother motion and acceleration profiles during crane manipulation.

Grinding, i.e. reducing the particle size of mined ore, is often the bottleneck of the mining concentrating process. Thus, even small improvements may lead to large increases in profit. The goal of the grinding circuit is two-sided; to maximize the throughput of ore, and minimize the resulting particle size of the ground ore within some acceptable range. In this work we study the control of a two-stage grinding circuit using reinforcement learning. To this end, we present a solution for integrating industrial simulation models into the reinforcement learning framework OpenAI Gym.We compare an existing PID controller, based on vast domain knowledge and years of hand-tuning, with a black-box algorithm called Proximal Policy Optimization on a calibrated grinding circuit simulation model. The comparison show that it is possible to control the grinding circuit using reinforcement learning. In addition, contrasting reinforcement learning from the existing PID control, the algorithm is able to maximize an abstract control goal: maximizing profit as defined by a profit function given by our industrial collaborator. In some operating cases the algorithm is able to control the plant more efficiently compared to existing control.

We present a fluid simulation method based on Smoothed Particle Hydrodynamics (SPH) in which incompressibility and boundary conditions are enforced using holonomic kinematic constraints on the density. This formulation enables systematic multiphysics integration in which interactions are modeled via similar constraints between the fluid pseudoparticles and impenetrable surfaces of other bodies. These conditions embody Archimede's principle for solids and thus buoyancy results as a direct consequence. We use a variational time stepping scheme suitable for general constrained multibody systems we call SPOOK. Each step requires the solution of only one Mixed Linear Complementarity Problem (MLCP) with very few inequalities, corresponding to solid boundary conditions. We solve this MLCP with a fast iterative method. Overall stability is vastly improved in comparison to the unconstrained version of SPH, and this allows much larger time steps, and an increase in overall performance by two orders of magnitude. Proof of concept is given for computer graphics applications and interactive simulations.

We describe a method for the visual interactive simulation of wires contacting with rigid multibodies. The physical model used is a hybrid combining lumped elements and massless quasistatic representations. The latter is based on a kinematic constraint preserving the total length of the wire along a segmented path which can involve multiple bodies simultaneously and dry frictional contact nodes used for roping, lassoing, and fastening. These nodes provide stick and slide friction along the edges of the contacting geometries. The lumped element resolution is adapted dynamically based on local stability criteria, becoming coarser as the tension increases, and up to the purely kinematic representation. Kinematic segments and contact nodes are added, deleted, and propagated based on contact geometries and dry friction configurations. The method gives a dramatic increase in both performance and robustness because it quickly decimates superfluous nodes without loosing stability, yet adapts to complex configurations with many contacts and high curvature, keeping a fixed, large integration time step. Numerical results demonstrating the performance and stability of the adaptive multiresolution scheme are presented along with an array of representative simulation examples illustrating the versatility of the frictional contact model.

The present paper addresses real-time simulation of cables for virtual environments. A faithful physical model based on constrained rigid bodies is introduced and discretized. The performance and stability of the numerical method are analyzed in details and found to meet the requirements of interactive heavy hoisting simulations. The physical model is well behaved in the limit of infinite stiffness as well as in the elastic regime, and the tuning parameters correspond directly to conventional material constants. The integration scheme mixes the well known Störmer-Verlet method for the dynamics equations with the linearly implicit Euler method for the constraint equations and enables physical constraint relaxation and stabilization terms. The technique is shown to have superior numerical stability properties in comparison with either chain link systems, or spring and damper models. Experimental results are presented to show that the method results in stable, real-time simulations. Stability persists for moderately large fixed integration step of Delta t = 1/60 s, with hoisting loads of up to 10(5) times heavier than the elements of the cable. Further numerical experiments validating the physical model are also presented.

Abstract A technique for real‐time simulation of hoisting cable systems based on a multibody nonideal constraint is presented. The hoisting cable constraint is derived from the cable internal energies for stretching and twisting. Each hoisting cable introduces two constraint equations, one for stretching and one for torsion, which include all the rigid bodies attached by the same cable. The computation produces the global tension and torsion in the cable as well as the resulting forces and torques on each attached body. The complexity of the computation grows linearly with the number of bodies attached to a given cable and is weakly coupled to the rest of the simulation. The nonideal constraint formulation allows stable simulations of cables over wide ranges of linear and torsional stiffness, including the rigid limit. This contrasts with lumped element formulations including the cable internal degrees of freedom in which computational complexity grows at least linearly with the number of cable elements – usually proportional to cable length – and where numerical stability is sensitive to the mass ratio between the load and the lumped elements.