For obvious reasons, conditions on lunar and planetary surfaces can understandably be hard to duplicate physically, even in the best case when they are “well understood,” if only from a distance. This makes digital simulations an essential tool during mission planning and development. Simulation can be used to study the effects of changes in terrain, lighting, reflectance, and other environmental factors on mission success.
In a typical space mission, though, many conditions are poorly understood. There may, for example, be limited understanding of soil—or regolith—properties before contact on a lunar mission. Addressing this requires special conceptual and mathematical tools. Parameters must be randomized to capture potential outcomes, and results must be optimized to discover the corner cases and unexpected outcomes that could impact a mission. Energid Technologies Corp. has been funded by NASA to develop a new robot simulation that accommodates uncertainty and discovers exceptional behaviors during mission planning.
"Our approach is to apply concepts from game theory and stochastic optimization to deeply simulate NASA's robotic missions," said Ryan Penning, Project Manager on the program. "The result will be a breakthrough ability to reason about uncertain environments and understand the extremes of what a robot can do." Energid Technologies specializes in the control, simulation, and sensing of complex systems for the aerospace, agriculture, transportation, defense, and medical industries.
Energid, through its Actin software, brings enabling tools and capability to the project. Actin supports randomizing simulations of all types of robotic systems. It has physics-based models for articulated dynamics, contact dynamics, sensor simulation, and communications. In this effort, Actin, and its stochastic simulation capability, will be extended to specialize this capability for new NASA space applications.
Some of Actin’s characteristics that make it particularly suited for such applications include it automated reaction to dynamic workspaces, including collision avoidance; finely scripted velocity profiles of manipulated objects; and complex dexterous manipulation. Actin is built on what Energid says is an “industry-leading” proprietary optimization codes that enables the robots that it controls to run some of the most challenging tasks in robotics today, all within a drag-and-drop interface.
Actin will be tailored to the space environment by modeling lunar regolith with highly parallelized particle models implemented on graphical processing units (GPUs). GPUs allow execution of high-fidelity simulation in real time on common computer hardware. Actin will also be tailored to support the appearance of lunar and planetary surfaces. This will allow high fidelity simulation of cameras and other sensors.
"Actin already provides powerful tools for camera, lidar, radar, and other sensors on all types of robots, from oil exploration to collaborative manufacturing," said James English, CTO at Energid. "This project will enable high-fidelity extensions for space environments." English told Aerospace Engineering that the work will focus primarily on NASA’s KREX platform.
The technology will be further commercialized by applying it to configure robotic systems and workcells on Earth. There is a pressing need for easier robotic programming to lower costs and empower people and businesses untrained in robotics but familiar with application domains where robots can contribute.
"Much of the cost of applying robots lies in configuring environments and workcells," said Neil Tardella, CEO at Energid. "The prediction and simulation technology developed under this project will lower the cost of fielding robots and expand their application."
Work on the project will be done in Wisconsin and Massachusetts.