No abstract

The reinforcement learning control of an underground loader was investigated in a simulated environment by using a multi-agent deep neural network approach. At the start of each loading cycle, one agent selects the dig position from a depth camera image of a pile of fragmented rock. A second agent is responsible for continuous control of the vehicle, with the goal of filling the bucket at the selected loading point while avoiding collisions, getting stuck, or losing ground traction. This relies on motion and force sensors, as well as on a camera and lidar. Using a soft actor–critic algorithm, the agents learn policies for efficient bucket filling over many subsequent loading cycles, with a clear ability to adapt to the changing environment. The best results—on average, 75% of the max capacity—were obtained when including a penalty for energy usage in the reward.

No abstract

PreprintThis is the submitted version of a paper published in Robotics and Autonomous Systems.Citation for the original published paper (version of record):Lindmark, D M., Servin, M. (2018) Computational exploration of robotic rock loading Robotics and Autonomous SystemsAbstract A method for simulation-based development of robotic rock loading systems is described and tested. The idea is to first formulate a generic loading strategy as a function of the shape of the rock pile, the kinematics of the machine and a set of motion design variables that will be used by the autonomous control system. The relation between the loading strategy and resulting performance is then explored systematically using contacting multibody dynamics simulation, multiobjective optimisation and surrogate modelling. With the surrogate model it is possible to find Pareto optimal loading strategies for dig plans that are adapted to the current shape of the pile. The method is tested on a load-haul-dump machine loading from a large muck pile in an underground mine, with the loading performance measured by productivity, machine wear and rock debris spill that cause interruptions. 45 model includes actuator models and a control algorithm to execute the motion plan. The actuators have limited force, and the machine may lose ground traction if the pile resistance is large enough to cause frictional slippage. Consequently, the simulated bucket trajectory may deviate from the plan. This 50 provides realism that is not available when using a kinematic model of the vehicle or bucket. For each simulation, the forces and moments in the joints of the machine are measured, as well as the loading cycle time, final bucket fill ratio as well as the resulting state of the pile. Both the actuating forces and the 55 constraint forces keeping the joints in place are measured, and may be analysed for power consumption and mechanical wear, respectively.A surrogate model is constructed for the relationships between the design variables and the objective functions, i.e., be-60 tween the loading strategy and loading performance. This enables systematic exploration of the design space and the possibility to steer the computational resources to the most interesting parts of the design space. Using the surrogate model, it is easy to answer specific questions regarding loading strate-65

History as a school subject has been a thorny issue for advocates of peace education at least since the 1880s. Efforts, including the substitution of cultural history for military history, have been made to ensure that history teaching promotes international understanding, not propagates chauvinism. The Norden Associations of Scandinavia, which were involved in textbook revision since 1919, achieved some success by altering contents, but national myths remained central to each country's historical narrative, making it difficult to give history education its desired international orientation.